Video wird geladen...

Video konnte nicht geladen werden

Beim Laden dieses Videos ist ein Problem aufgetreten. Dies könnte an einem vorübergehenden Netzwerkproblem liegen oder das Video ist möglicherweise nicht verfügbar.

I’ve been testing Hy4 preview in WorkBuddy, and the most interesting part is not simply the model size, it’s how much practical work it can handle with a relatively focused active parameter count. Hy4 preview brings together stronger code understanding, generation, and editing; improved document and information processing; workflow... automation; web and game development; cross-tool collaboration; and more reliable completion of complex, multi-step tasks. In other words, it is designed for work that requires planning, tool use, iteration, and follow-through, not just a quick answer in a chat window. Compared with its initial release, the current Hy4 preview is noticeably faster and better-performing in practical workflows. Following an upgrade released yesterday, it can complete tasks with fewer conversation rounds and lower token usage, while reasoning more quickly and making the overall user experience feel smoother from the first instruction to the final result. For my test, I gave it a demanding Three.js game-prototyping task with a 770B-parameter model and 49B active parameters. The result was more revealing than a simple first-look demo: Hy4 preview handled the core logic, edge cases, and follow-up changes while maintaining the broader context of the project. That combination of capability, speed, context, and active compute is what makes its cost-effectiveness worth examining. A fair evaluation should use the same prompt and environment configuration across models, changing only the model itself. That makes it easier to assess task completion, planning quality, tool-calling stability, reasoning speed, token efficiency, and performance over longer workflows without confusing the result with different settings. If you want to test the model yourself, access Hy4 preview through WorkBuddy and see how it performs on a real coding, document, automation, or creative task: Tencent Hy Tencent AI WorkBuddyshow more

Tyler Wayne

407,183 subscribers

56,152 Aufrufe • vor 7 Tagen •via X (Twitter)

Bildung Wissenschaft & Technologie

Anya Rossi• Live Now

Private livecam show

25 Kommentare

Profilbild von MAUREEN

MAUREENvor 6 Tagen

A capable workflow should support iteration without losing the original goal.

Profilbild von Stones

Stonesvor 6 Tagen

Practical coding support is about solving real problems, not only producing output.

Profilbild von Bonita🤎

Bonita🤎vor 6 Tagen

Code generation becomes more useful when editing is part of the same process.

Profilbild von RICHIE..😎

RICHIE..😎vor 6 Tagen

Developers need tools that can follow the context of an evolving project.

Profilbild von SHERIFF™🤠

SHERIFF™🤠vor 6 Tagen

Real workflow testing reveals more than a benchmark summary ever can.

Profilbild von George

Georgevor 6 Tagen

It helps when a model can contribute across planning, writing, and revision.

Profilbild von JA KISII™ 🇰🇪

JA KISII™ 🇰🇪vor 6 Tagen

Focused active capacity can be valuable for everyday coding tasks.

Profilbild von Tyla

Tylavor 6 Tagen

Developers need systems that can work with context instead of ignoring it.

Profilbild von JOMBA

JOMBAvor 6 Tagen

A useful coding assistant should help maintain momentum across tasks.

Profilbild von CATHEY

CATHEYvor 6 Tagen

Better code support lets teams spend more time on the harder decisions.

Profilbild von Betty

Bettyvor 6 Tagen

The workflow matters as much as the model behind it.

Profilbild von Ochora🇰🇪☆, KC

Ochora🇰🇪☆, KCvor 6 Tagen

Practical usefulness matters more than scale when work needs to get done.

Profilbild von CATHEY

CATHEYvor 6 Tagen

Real productivity comes from fewer interruptions during the building process.

Profilbild von Kiage Clinton, KC

Kiage Clinton, KCvor 6 Tagen

Editing support matters because first drafts rarely stay unchanged.

Profilbild von iano

ianovor 6 Tagen

The best tools help turn a rough idea into something workable.

Profilbild von chenly🥹

chenly🥹vor 6 Tagen

Generating and refining should feel like parts of one process.

Profilbild von Jayden Cole

Jayden Colevor 6 Tagen

Practical performance is what determines whether a model earns repeat use.

Profilbild von Amazing

Amazingvor 6 Tagen

Strong understanding can reduce the back and forth during technical work.

Profilbild von Fatuma

Fatumavor 6 Tagen

Good tools reduce the effort required to move from draft to improvement.

Profilbild von Weigo

Weigovor 6 Tagen

A model should be judged by the quality of work it helps complete.

Profilbild von JADUONG

JADUONGvor 6 Tagen

Focused capability can make a big difference in regular product work. .

Profilbild von JADUONG

JADUONGvor 6 Tagen

Code understanding matters when a task involves more than writing new files.

Profilbild von Alex Ryan

Alex Ryanvor 6 Tagen

Model size only matters if it translates into useful results.

Profilbild von REX JAMES

REX JAMESvor 6 Tagen

A model becomes more helpful when it can revise work thoughtfully.

Profilbild von Saeed Anwar

Saeed Anwarvor 6 Tagen

Fewer rounds and lower token usage matter more than benchmark scores in production agent cost. What tasks did you actually run it on?

Ähnliche Videos

I gave an AI one instruction: “Build me a playable 3D game from scratch.” Then I let it cook. 👀 I tested Tencent Hy4 preview inside WorkBuddy, and it went way beyond just generating a few snippets of code. It could: → Plan the game structure → Generate the code across files → Create the UI and interactions → Debug problems along the way → Iterate on the project while keeping context That last part is what caught my attention. Hy4 preview is built for these long-horizon tasks, with 770B total parameters, 49B active parameters, and a 1M-token context window. It’s designed for coding agents, game development, tool use, research, and other workflows where the model has to actually do the work, not just answer a prompt. And right now, you can try Hy4 preview for free on WorkBuddy until September 10. If you want to see how far an AI agent can take a real project, try it yourself: See what you can build with it. Tencent AI

I gave an AI one instruction: “Build me a playable 3D game from scratch.” Then I let it cook. 👀 I tested Tencent Hy4 preview inside WorkBuddy, and it went way beyond just generating a few snippets of code. It could: → Plan the game structure → Generate the code across files → Create the UI and interactions → Debug problems along the way → Iterate on the project while keeping context That last part is what caught my attention. Hy4 preview is built for these long-horizon tasks, with 770B total parameters, 49B active parameters, and a 1M-token context window. It’s designed for coding agents, game development, tool use, research, and other workflows where the model has to actually do the work, not just answer a prompt. And right now, you can try Hy4 preview for free on WorkBuddy until September 10. If you want to see how far an AI agent can take a real project, try it yourself: See what you can build with it. Tencent AI

Toha Khan

66,214 Aufrufe • vor 12 Tagen

🚨 Tencent just dropped a serious open-source AI contender. Meet Hy4 Preview built for productivity, coding & agentic workflows. 770B total parameters. 49B active. 1M+ token context. But the numbers aren’t the most interesting part. Hy4 Preview is designed to handle what actually matters: planning, tool use, coding, research, document workflows and complex multi-step tasks. In Tencent’s internal blind evaluation across 203 engineering tasks, it scored 2.99/4, edging out GLM 5.3 (2.92) and Kimi K3 (2.94). From full-stack development and research to workplace automation and even game development, this looks less like “just another LLM” and more like a model built to get real work done. Worth watching 👀 🔗 Try it with WorkBuddy: Tencent Hy Tencent AI WorkBuddy

🚨 Tencent just dropped a serious open-source AI contender. Meet Hy4 Preview built for productivity, coding & agentic workflows. 770B total parameters. 49B active. 1M+ token context. But the numbers aren’t the most interesting part. Hy4 Preview is designed to handle what actually matters: planning, tool use, coding, research, document workflows and complex multi-step tasks. In Tencent’s internal blind evaluation across 203 engineering tasks, it scored 2.99/4, edging out GLM 5.3 (2.92) and Kimi K3 (2.94). From full-stack development and research to workplace automation and even game development, this looks less like “just another LLM” and more like a model built to get real work done. Worth watching 👀 🔗 Try it with WorkBuddy: Tencent Hy Tencent AI WorkBuddy

RAVI KUMAR SAHU

63,568 Aufrufe • vor 18 Tagen

The next big advantage in AI might not come from one breakthrough. It might come from how quickly a team can build, learn, and improve. Tencent Hunyuan has now released three major models in six months. Hy4 preview is the latest result of that pace. It is fully open source, with 770B total parameters, 49B active parameters, and support for more than 1M tokens of context. The model is designed around practical work, including coding, office productivity, science, planning, and longer workflows that require sustained execution. Access is priced at $0.834 per million input tokens, $2.501 per million output tokens, and $0.042 per million cache hits. That combination of rapid iteration, open access, and affordable usage is what makes this release worth paying attention to. Hy4 preview can be experienced free on WorkBuddy for the next two weeks. Tencent Hy Tencent AI WorkBuddy

The next big advantage in AI might not come from one breakthrough. It might come from how quickly a team can build, learn, and improve. Tencent Hunyuan has now released three major models in six months. Hy4 preview is the latest result of that pace. It is fully open source, with 770B total parameters, 49B active parameters, and support for more than 1M tokens of context. The model is designed around practical work, including coding, office productivity, science, planning, and longer workflows that require sustained execution. Access is priced at $0.834 per million input tokens, $2.501 per million output tokens, and $0.042 per million cache hits. That combination of rapid iteration, open access, and affordable usage is what makes this release worth paying attention to. Hy4 preview can be experienced free on WorkBuddy for the next two weeks. Tencent Hy Tencent AI WorkBuddy

Manish Kumar Shah

42,263 Aufrufe • vor 18 Tagen

🚨 Tencent Hunyuan just dropped Hy4 preview—and it’s already making its way into the open-source big leagues. Three major releases in six months. That’s seriously fast. Hy4 preview comes with: 770B total parameters 49B active parameters A 1M+ token context window And it’s built with one clear focus: productivity. We’re talking coding, office work, science, software engineering, gaming, finance, security, and more. But here’s the part that really caught my attention: Hy4 preview wasn’t developed in isolation. It was co-designed with real products like WorkBuddy. Tencent worked with experts from across its ecosystem to build the training data, then tested the model on real-world tasks. In a blind test involving 163 internal experts and 203 engineering tasks, the scores were: Hy4 preview: 2.99/4 Kimi K3: 2.94/4 GLM 5.3: 2.92/4 And performance isn’t the only impressive part. On benchmarks, Hy4 preview clearly outperforms GLM 5.2 and is basically neck-and-neck with GLM 5.3. Then there’s the pricing: $6/M input tokens $18/M output tokens $0.30/M cache hits That’s flagship-level performance without the flagship-level price. You can try Hy4 preview for FREE through WorkBuddy over the next two weeks: Tencent Hy @TencentAl_News

🚨 Tencent Hunyuan just dropped Hy4 preview—and it’s already making its way into the open-source big leagues. Three major releases in six months. That’s seriously fast. Hy4 preview comes with: 770B total parameters 49B active parameters A 1M+ token context window And it’s built with one clear focus: productivity. We’re talking coding, office work, science, software engineering, gaming, finance, security, and more. But here’s the part that really caught my attention: Hy4 preview wasn’t developed in isolation. It was co-designed with real products like WorkBuddy. Tencent worked with experts from across its ecosystem to build the training data, then tested the model on real-world tasks. In a blind test involving 163 internal experts and 203 engineering tasks, the scores were: Hy4 preview: 2.99/4 Kimi K3: 2.94/4 GLM 5.3: 2.92/4 And performance isn’t the only impressive part. On benchmarks, Hy4 preview clearly outperforms GLM 5.2 and is basically neck-and-neck with GLM 5.3. Then there’s the pricing: $6/M input tokens $18/M output tokens $0.30/M cache hits That’s flagship-level performance without the flagship-level price. You can try Hy4 preview for FREE through WorkBuddy over the next two weeks: Tencent Hy @TencentAl_News

AGAFE

131,568 Aufrufe • vor 18 Tagen

🚨 I just built a game with an open-source AI model. And honestly… I didn’t expect it to be this capable. Tencent Hunyuan just released Hy4 preview, and it’s already pushing into the top tier of open-source models. Three major releases in six months. That pace is crazy. Here’s what Hy4 preview brings: → 770B total parameters → 49B active parameters → 1M+ token context window → Fully open-source But the numbers aren’t even the most interesting part. Hy4 preview was built around one goal: real-world productivity. Coding. Engineering. Office work. Science. Gaming. Finance. Security. And Tencent didn’t build it in isolation. Hy4 preview was co-designed alongside real products like WorkBuddy, using expertise and real-world data from across Tencent’s ecosystem. So I decided to test it the way I actually like testing AI models: I gave it a game idea and let WorkBuddy help turn it into a playable experience. 🎮 From the initial concept to the actual game logic, it was surprisingly smooth. And the benchmark results back up the hype: 163 internal experts 203 engineering tasks Hy4 preview — 2.99/4 Kimi K3 — 2.94/4 GLM 5.3 — 2.92/4 It also beats GLM 5.2 on benchmarks and comes remarkably close to GLM 5.3. Then comes the part I really like: 💰 ¥6/M input tokens 💰 ¥18/M output tokens 💰 ¥0.30/M cache hits Flagship-level capability without the flagship-level price. And right now, you can try Hy4 preview FREE through WorkBuddy for the next two weeks. If you’re curious what it can actually do, don’t just read the benchmarks. Build something with it. 🔗 Tencent Hy Tencent AI

🚨 I just built a game with an open-source AI model. And honestly… I didn’t expect it to be this capable. Tencent Hunyuan just released Hy4 preview, and it’s already pushing into the top tier of open-source models. Three major releases in six months. That pace is crazy. Here’s what Hy4 preview brings: → 770B total parameters → 49B active parameters → 1M+ token context window → Fully open-source But the numbers aren’t even the most interesting part. Hy4 preview was built around one goal: real-world productivity. Coding. Engineering. Office work. Science. Gaming. Finance. Security. And Tencent didn’t build it in isolation. Hy4 preview was co-designed alongside real products like WorkBuddy, using expertise and real-world data from across Tencent’s ecosystem. So I decided to test it the way I actually like testing AI models: I gave it a game idea and let WorkBuddy help turn it into a playable experience. 🎮 From the initial concept to the actual game logic, it was surprisingly smooth. And the benchmark results back up the hype: 163 internal experts 203 engineering tasks Hy4 preview — 2.99/4 Kimi K3 — 2.94/4 GLM 5.3 — 2.92/4 It also beats GLM 5.2 on benchmarks and comes remarkably close to GLM 5.3. Then comes the part I really like: 💰 ¥6/M input tokens 💰 ¥18/M output tokens 💰 ¥0.30/M cache hits Flagship-level capability without the flagship-level price. And right now, you can try Hy4 preview FREE through WorkBuddy for the next two weeks. If you’re curious what it can actually do, don’t just read the benchmarks. Build something with it. 🔗 Tencent Hy Tencent AI

Aryan Rakib

63,055 Aufrufe • vor 6 Tagen

Tencent just released Hy4 Preview: a 770B-parameter open-source model with 49B active parameters and a 1M-token context window. The most interesting part is its agentic research capability. Tencent says Hy4 coordinated several Codex sessions in parallel, evaluated their results and adjusted the research direction, outperforming Codex working alone across eight benchmarks. Hy4 is aimed at long-horizon coding, multi-file office work and scientific research. In sum: a solid, good release. A few months ago this would be absolute sota.

Tencent just released Hy4 Preview: a 770B-parameter open-source model with 49B active parameters and a 1M-token context window. The most interesting part is its agentic research capability. Tencent says Hy4 coordinated several Codex sessions in parallel, evaluated their results and adjusted the research direction, outperforming Codex working alone across eight benchmarks. Hy4 is aimed at long-horizon coding, multi-file office work and scientific research. In sum: a solid, good release. A few months ago this would be absolute sota.

Chubby♨️

49,575 Aufrufe • vor 20 Tagen

I wanted to see what Hy4 preview could actually do when I gave it a real creative project instead of another “write me some code” prompt. So I went with something I’d genuinely want to play: 🐬 a cute dolphin as the hero 🌊 colorful underwater worlds 💎 glowing pearls to collect 🐠 sea creatures and obstacles 🪸 coral reefs, bubbles & hidden treasure 🎮 smoother animations and progressively harder levels The result is this underwater platformer 👇 And this is where Hy4 preview gets interesting to me. Behind the scenes, we’re talking about 770B parameters, 49B active parameters and a 1M+ token context window. It’s also fully open-sourced. But big numbers only matter if they translate into something useful. What I’m really testing is how much actual work the model can take off my plate understanding the idea, handling the implementation, keeping all the moving parts connected, and getting from a blank canvas to something that feels like a real experience. And there’s another update that makes this even more interesting: Hy4 preview was upgraded yesterday to significantly reduce both conversation turns and token consumption. That means less back-and-forth, faster thinking, and a smoother overall experience when working through complex tasks. That’s also why the cost-effectiveness angle stands out to me. If a model can handle more of the workflow while using fewer tokens and requiring fewer iterations, the value isn’t just in the benchmark score. It’s in how much more you can actually build with it. This dolphin has me wanting to test Hy4 preview with some much crazier ideas next. 🐬👀 Tencent Hy Tencent AI ☞

I wanted to see what Hy4 preview could actually do when I gave it a real creative project instead of another “write me some code” prompt. So I went with something I’d genuinely want to play: 🐬 a cute dolphin as the hero 🌊 colorful underwater worlds 💎 glowing pearls to collect 🐠 sea creatures and obstacles 🪸 coral reefs, bubbles & hidden treasure 🎮 smoother animations and progressively harder levels The result is this underwater platformer 👇 And this is where Hy4 preview gets interesting to me. Behind the scenes, we’re talking about 770B parameters, 49B active parameters and a 1M+ token context window. It’s also fully open-sourced. But big numbers only matter if they translate into something useful. What I’m really testing is how much actual work the model can take off my plate understanding the idea, handling the implementation, keeping all the moving parts connected, and getting from a blank canvas to something that feels like a real experience. And there’s another update that makes this even more interesting: Hy4 preview was upgraded yesterday to significantly reduce both conversation turns and token consumption. That means less back-and-forth, faster thinking, and a smoother overall experience when working through complex tasks. That’s also why the cost-effectiveness angle stands out to me. If a model can handle more of the workflow while using fewer tokens and requiring fewer iterations, the value isn’t just in the benchmark score. It’s in how much more you can actually build with it. This dolphin has me wanting to test Hy4 preview with some much crazier ideas next. 🐬👀 Tencent Hy Tencent AI ☞

Md Riyazuddin

63,562 Aufrufe • vor 8 Tagen

I wanted to see where Hy4 preview would stop, so I gave it a deliberately demanding prompt: “Build a working Cyber Incident Command Center.” Not a mockup. A real, responsive app with a live threat map, animated attack routes, incident simulation, filters, investigation panels, event logs, ownership controls, and working isolate/resolve actions. Hy4 preview planned the build, wrote the application, ran it locally, tested the interactions, checked the mobile layout, fixed issues, and opened the finished product for review. Then I clicked “Simulate.” A new incident appeared, the dashboard updated, I filtered it by severity, opened the evidence trail, isolated the affected system, and resolved it. The KPI counters changed with the action. That’s a much more useful coding demo to me than watching code appear in an editor. Hy4 preview is Hunyuan’s third major release in six months. It’s fully open-sourced, with 770B total parameters, 49B activated, and a 1M+ token context window. Tencent built it around productivity scenarios through close work with experts in software engineering, gaming, finance, and security, plus direct co-design with tools like WorkBuddy. In an internal blind test covering 203 engineering tasks and 163 Tencent experts, it scored 2.99/4—slightly ahead of Kimi K3 at 2.94 and GLM 5.3 at 2.92. And the pricing stays practical: $0.834/M input $2.501/M output $0.042/M cache hits If you want to test it yourself, Hy4 preview is free inside WorkBuddy for the next two weeks, for a limited time. Tencent Hy Tencent AI WorkBuddy

I wanted to see where Hy4 preview would stop, so I gave it a deliberately demanding prompt: “Build a working Cyber Incident Command Center.” Not a mockup. A real, responsive app with a live threat map, animated attack routes, incident simulation, filters, investigation panels, event logs, ownership controls, and working isolate/resolve actions. Hy4 preview planned the build, wrote the application, ran it locally, tested the interactions, checked the mobile layout, fixed issues, and opened the finished product for review. Then I clicked “Simulate.” A new incident appeared, the dashboard updated, I filtered it by severity, opened the evidence trail, isolated the affected system, and resolved it. The KPI counters changed with the action. That’s a much more useful coding demo to me than watching code appear in an editor. Hy4 preview is Hunyuan’s third major release in six months. It’s fully open-sourced, with 770B total parameters, 49B activated, and a 1M+ token context window. Tencent built it around productivity scenarios through close work with experts in software engineering, gaming, finance, and security, plus direct co-design with tools like WorkBuddy. In an internal blind test covering 203 engineering tasks and 163 Tencent experts, it scored 2.99/4—slightly ahead of Kimi K3 at 2.94 and GLM 5.3 at 2.92. And the pricing stays practical: $0.834/M input $2.501/M output $0.042/M cache hits If you want to test it yourself, Hy4 preview is free inside WorkBuddy for the next two weeks, for a limited time. Tencent Hy Tencent AI WorkBuddy

Clara Bennett

12,864 Aufrufe • vor 17 Tagen

tencent hunyuan just shipped their third major model in six months. that pace is hard to keep up with. Hy4 preview is now live in workbuddy, and it's fully open source. the specs: 770B total params, 49B active, 1M+ token context window. but the number that matters is where it's pointed — real work. coding, office tasks, research, documents. here's the part i find interesting. tencent didn't build this in isolation. they put the training data together with their own experts across software engineering, gaming, finance and security, then designed the model side by side with products like workbuddy. model and product shaping each other instead of one being bolted onto the other. and it shows up in testing. 163 internal experts ran a blind test across 203 engineering tasks. hy4 preview scored 2.99/4 — ahead of kimi k3 at 2.94 and glm 5.3 at 2.92. on public benchmarks it clearly beats glm 5.2 and sits about level with glm 5.3. then the pricing. $6 per million input tokens, $18 per million output, $0.3 per million on cache hits. flagship results at a price you don't have to think about. and for the next two weeks it's free inside workbuddy.

tencent hunyuan just shipped their third major model in six months. that pace is hard to keep up with. Hy4 preview is now live in workbuddy, and it's fully open source. the specs: 770B total params, 49B active, 1M+ token context window. but the number that matters is where it's pointed — real work. coding, office tasks, research, documents. here's the part i find interesting. tencent didn't build this in isolation. they put the training data together with their own experts across software engineering, gaming, finance and security, then designed the model side by side with products like workbuddy. model and product shaping each other instead of one being bolted onto the other. and it shows up in testing. 163 internal experts ran a blind test across 203 engineering tasks. hy4 preview scored 2.99/4 — ahead of kimi k3 at 2.94 and glm 5.3 at 2.92. on public benchmarks it clearly beats glm 5.2 and sits about level with glm 5.3. then the pricing. $6 per million input tokens, $18 per million output, $0.3 per million on cache hits. flagship results at a price you don't have to think about. and for the next two weeks it's free inside workbuddy.

Fakhr

55,316 Aufrufe • vor 19 Tagen

Hy4 preview just dropped, so I tested it on WorkBuddy against three other models: DeepSeek V4 Flash, GPT-5.6-Luna, and GLM-5.3... the top 4 models on this week’s OpenRouter coding leaderboard. The best part? They’re all available in WorkBuddy, so you don’t need to download multiple Agent apps. For the test, I used one single prompt to build one of those games you’ve definitely seen in ads: swipe/move through number gates, make calculations, and eventually defeat the boss. The results were pretty interesting 👇 For a task of this difficulty, everyone performed well except DeepSeek. It seemed to finish coding without checking for bugs, and the game couldn’t run properly. GPT-5.6-Luna worked, but felt like an arithmetic multiple-choice quiz. A little boring. GLM-5.3 and Hy4 preview were much more impressive. GLM was slower (video sped up 2.2×) and only had three movement positions, while Hy4 offered continuous horizontal movement and a more natural speed. What surprised me most was Hy4’s game design: besides basic arithmetic, it added weapons that reduce enemy damage or increase stolen power, making the game much more fun. Also worth noting: Tencent’s Sherry quantization compresses the 770B MoE model from ~1.5TB BF16 to 214GB, while keeping performance close to the original... making local deployment much more practical. Hy4 preview is free on WorkBuddy until September 10. Give it a try 👇

Hy4 preview just dropped, so I tested it on WorkBuddy against three other models: DeepSeek V4 Flash, GPT-5.6-Luna, and GLM-5.3... the top 4 models on this week’s OpenRouter coding leaderboard. The best part? They’re all available in WorkBuddy, so you don’t need to download multiple Agent apps. For the test, I used one single prompt to build one of those games you’ve definitely seen in ads: swipe/move through number gates, make calculations, and eventually defeat the boss. The results were pretty interesting 👇 For a task of this difficulty, everyone performed well except DeepSeek. It seemed to finish coding without checking for bugs, and the game couldn’t run properly. GPT-5.6-Luna worked, but felt like an arithmetic multiple-choice quiz. A little boring. GLM-5.3 and Hy4 preview were much more impressive. GLM was slower (video sped up 2.2×) and only had three movement positions, while Hy4 offered continuous horizontal movement and a more natural speed. What surprised me most was Hy4’s game design: besides basic arithmetic, it added weapons that reduce enemy damage or increase stolen power, making the game much more fun. Also worth noting: Tencent’s Sherry quantization compresses the 770B MoE model from ~1.5TB BF16 to 214GB, while keeping performance close to the original... making local deployment much more practical. Hy4 preview is free on WorkBuddy until September 10. Give it a try 👇

Orikan

40,312 Aufrufe • vor 13 Tagen

Codex can now run Hy4 preview! Tencent dropped it today. 770b total but only 49b active per token, 1m context, apache 2.0. A frontier model you can actually download. Their own blind eval put it ahead of GLM 5.3 and Kimi k3 on engineering work, 163 experts scoring 203 tasks. 65.7 on swe bench pro. Hy4 now sits in the picker next to sol 5.6. OpenCode and Command code already serves it on the $10 plan, so no api key, no per token metering, and my native models never moved. Here is how you set it up:

Codex can now run Hy4 preview! Tencent dropped it today. 770b total but only 49b active per token, 1m context, apache 2.0. A frontier model you can actually download. Their own blind eval put it ahead of GLM 5.3 and Kimi k3 on engineering work, 163 experts scoring 203 tasks. 65.7 on swe bench pro. Hy4 now sits in the picker next to sol 5.6. OpenCode and Command code already serves it on the $10 plan, so no api key, no per token metering, and my native models never moved. Here is how you set it up:

Ziwen

18,451 Aufrufe • vor 19 Tagen

Tencent just dropped Hy4 preview, and it's officially open source now. 770B total params, 49B active, 1M+ token context. three major releases in six months, and this one just landed in the top tier of open source models. → Built through Tencent's co-design loop: training data shaped with experts across software engineering, gaming, finance, and security, then built directly alongside products like WorkBuddy → In a blind test with 163 internal experts across 203 engineering tasks, it scored 2.99/4, ahead of GLM 5.3 at 2.92 and Kimi K3 at 2.94 → Benchmarks put it clearly past GLM 5.2 and neck and neck with GLM 5.3 → Pricing stayed low: $0.834/M input, $2.501/M output, $0.042/M cache hits I asked it to write a weekly status update email template for a team lead, then generate a filled-out example with placeholder projects, and export both as Word docs. one prompt, no follow-up, and it handed back both files clean and ready to send. free to try for the next two weeks through WorkBuddy: Tencent Hy Tencent AI WorkBuddy

Tencent just dropped Hy4 preview, and it's officially open source now. 770B total params, 49B active, 1M+ token context. three major releases in six months, and this one just landed in the top tier of open source models. → Built through Tencent's co-design loop: training data shaped with experts across software engineering, gaming, finance, and security, then built directly alongside products like WorkBuddy → In a blind test with 163 internal experts across 203 engineering tasks, it scored 2.99/4, ahead of GLM 5.3 at 2.92 and Kimi K3 at 2.94 → Benchmarks put it clearly past GLM 5.2 and neck and neck with GLM 5.3 → Pricing stayed low: $0.834/M input, $2.501/M output, $0.042/M cache hits I asked it to write a weekly status update email template for a team lead, then generate a filled-out example with placeholder projects, and export both as Word docs. one prompt, no follow-up, and it handed back both files clean and ready to send. free to try for the next two weeks through WorkBuddy: Tencent Hy Tencent AI WorkBuddy

D-Coder

109,151 Aufrufe • vor 18 Tagen

Tried Hy4 preview through WorkBuddy and gave it a pretty vague prompt for a first person roller coaster ride, front row seat, make it feel fast, that was basically it. No detailed spec, no reference video, just a rough idea in a sentence. What came back actually held up. Steep drop that felt like the bottom dropping out, banked turns that leaned the right way, scenery rushing by close enough that you register the speed instead of just watching it happen. For something built off a loose prompt like that, I did not expect it to nail the feel this well. Coding and 3D rendering have clearly gotten a real upgrade here, and this took one pass to look this good. Free to try in WorkBuddy for the next two weeks if you want to mess with it yourself. Tencent Hy Tencent AI

Tried Hy4 preview through WorkBuddy and gave it a pretty vague prompt for a first person roller coaster ride, front row seat, make it feel fast, that was basically it. No detailed spec, no reference video, just a rough idea in a sentence. What came back actually held up. Steep drop that felt like the bottom dropping out, banked turns that leaned the right way, scenery rushing by close enough that you register the speed instead of just watching it happen. For something built off a loose prompt like that, I did not expect it to nail the feel this well. Coding and 3D rendering have clearly gotten a real upgrade here, and this took one pass to look this good. Free to try in WorkBuddy for the next two weeks if you want to mess with it yourself. Tencent Hy Tencent AI

Chubby♨️

27,793 Aufrufe • vor 18 Tagen

CHINA JUST DROPPED AN AI CODING MODEL WITH A 1M CONTEXT WINDOW. And I connected it to Claude Code to see what it could actually do. Meet GLM-X Preview On paper, a few things immediately stood out: → 1M context window → Agentic coding capabilities → Works inside Claude Code → Designed for large-scale coding and reasoning workflows But specs don't matter much if the model can't deliver in practice. So I gave it a real-world task. THE TEST One prompt: > Build a modern AI lead generation dashboard using React and Tailwind CSS. Requirements: → Dark mode → Analytics dashboard → Lead table → Email outreach section → Responsive design → Production-ready component structure Instead of generating a few snippets, it planned the architecture, generated the dashboard components, created the Tailwind configuration, and walked through the implementation requirements. What impressed me most wasn't the code itself. It was how well it maintained context throughout the workflow. That's where a 1M context window starts becoming useful. Less time re-explaining requirements. Less context loss. More room for complex projects. The AI coding race is getting very interesting. And it's no longer just GPT, Claude, and Gemini competing for attention. Results from my test below 👇

CHINA JUST DROPPED AN AI CODING MODEL WITH A 1M CONTEXT WINDOW. And I connected it to Claude Code to see what it could actually do. Meet GLM-X Preview On paper, a few things immediately stood out: → 1M context window → Agentic coding capabilities → Works inside Claude Code → Designed for large-scale coding and reasoning workflows But specs don't matter much if the model can't deliver in practice. So I gave it a real-world task. THE TEST One prompt: > Build a modern AI lead generation dashboard using React and Tailwind CSS. Requirements: → Dark mode → Analytics dashboard → Lead table → Email outreach section → Responsive design → Production-ready component structure Instead of generating a few snippets, it planned the architecture, generated the dashboard components, created the Tailwind configuration, and walked through the implementation requirements. What impressed me most wasn't the code itself. It was how well it maintained context throughout the workflow. That's where a 1M context window starts becoming useful. Less time re-explaining requirements. Less context loss. More room for complex projects. The AI coding race is getting very interesting. And it's no longer just GPT, Claude, and Gemini competing for attention. Results from my test below 👇

Md Riyazuddin

31,199 Aufrufe • vor 3 Monaten

“Atria Dawn Preview” just caught my attention after I tested it on a real frontend task. I tested Atria ASI with Claude Code on a real frontend task: build a modern, responsive landing page using HTML, CSS, and JavaScript. It created the files, validated the code, and opened the final result for review. What stood out was the workflow. It went from understanding the task to writing code, checking the result, and delivering something I could actually run and verify. That makes “Atria Dawn Preview” interesting to me for research, development, and other professional workflows where the goal isn’t just an answer, but a usable and verifiable result. Definitely an agent I’ll be keeping an eye on. Official links:👇 API: Website: Looking forward to seeing where Atria Dawn Preview goes next.

“Atria Dawn Preview” just caught my attention after I tested it on a real frontend task. I tested Atria ASI with Claude Code on a real frontend task: build a modern, responsive landing page using HTML, CSS, and JavaScript. It created the files, validated the code, and opened the final result for review. What stood out was the workflow. It went from understanding the task to writing code, checking the result, and delivering something I could actually run and verify. That makes “Atria Dawn Preview” interesting to me for research, development, and other professional workflows where the goal isn’t just an answer, but a usable and verifiable result. Definitely an agent I’ll be keeping an eye on. Official links:👇 API: Website: Looking forward to seeing where Atria Dawn Preview goes next.

Aryan Rakib

35,129 Aufrufe • vor 2 Tagen

Progress in open models is keeping Big AI labs up at night, and I'm here for it! We have a brand new open-weight multimodal model optimized for long-horizon tasks. This model is really good at something: it can work on tasks that keep evolving over time. • 280B total parameters, but only 16B active • 512K context window • Understands text, images, and audio • Strong reasoning, coding, and tool use But the best of all: the model learns and adapts to new information! Imagine you start running an agent today to solve a problem, and while it's working, you get new information that changes the initial conditions, or you change your mind. The agents you run today don't have issues with short tasks and goals that don't change, but reality is messy, and that makes it hard for long-horizon agents to succeed. The new dots3-note Preview model introduces TEMPO. TEMPO is a new reinforcement learning technique that lets the model periodically pause and critique its own progress. Basically, from time to time, the agent asks itself: "Am I getting closer to the goal, or am I wasting my time?" The same model switches between actor and critic. The actor works on the problem. The critic looks at the current state, reasons about how much progress it has made, and determines what should happen next. TEMPO gives the model feedback along the way. This is huge for any agent that can work on long-horizon tasks without wasting its time.

Santiago

80,792 Aufrufe • vor 28 Tagen

I think we're reaching a point where context length, reasoning quality, and cost matter more than another leaderboard screenshot. After spending some time with Tencent Hunyuan Hy3, that's exactly the impression I got. The model handled long coding sessions, instruction following, and agent-style tasks without feeling unnecessarily "heavy." Instead of optimizing for one-off prompts, it feels designed for people who actually build with LLMs every day. Another thing that caught my attention is how much performance it packs into its architecture. According to Tencent, Hy3 delivers intelligence that's competitive with flagship open-source models 2–5× larger in parameter scale, making it an impressive balance of capability and efficiency. It's also released under the Apache 2.0 license, available through OpenRouter, and currently free to try for the next two weeks, making it a great opportunity for developers to experiment without worrying about inference costs. Curious to see how the community puts it to work over the next few weeks. Try it here: Thanks Tencent Hy for bringing another strong open model to developers. #Hy3 #Hunyuan #TencentAI #OpenSource

I think we're reaching a point where context length, reasoning quality, and cost matter more than another leaderboard screenshot. After spending some time with Tencent Hunyuan Hy3, that's exactly the impression I got. The model handled long coding sessions, instruction following, and agent-style tasks without feeling unnecessarily "heavy." Instead of optimizing for one-off prompts, it feels designed for people who actually build with LLMs every day. Another thing that caught my attention is how much performance it packs into its architecture. According to Tencent, Hy3 delivers intelligence that's competitive with flagship open-source models 2–5× larger in parameter scale, making it an impressive balance of capability and efficiency. It's also released under the Apache 2.0 license, available through OpenRouter, and currently free to try for the next two weeks, making it a great opportunity for developers to experiment without worrying about inference costs. Curious to see how the community puts it to work over the next few weeks. Try it here: Thanks Tencent Hy for bringing another strong open model to developers. #Hy3 #Hunyuan #TencentAI #OpenSource

Succeeded Mind

12,378 Aufrufe • vor 2 Monaten

Qwen3.8-Flash-Next is still going strong at 364.7K tokens of context on an M5 Max. And this isn’t just a static long-context test. The model was reasoning about how to speed up its own workflow while using tools, and the tool calls kept working without misses. Setup: • Qwen3.8-Flash-Next • M5 Max • 128GB unified memory • MLX-Serve PR #363 • OpenCode 2 • 364.7K context The interesting part isn’t simply getting hundreds of thousands of tokens into memory. It’s what happens once the context gets this large. Long-context inference usually comes with a painful tradeoff. As the KV cache grows, memory pressure increases and generation can slow down. But this setup is still pushing through 364K tokens while maintaining a usable agent workflow. The model can reason, call tools, inspect results, continue working, and keep the session moving. And the tool calls reportedly haven’t missed so far. That’s important for agentic coding. A huge context window is only useful if the model can actually operate reliably inside it. A 400K-token context that constantly breaks tool calls isn’t very useful. A 364K session that can keep reasoning and executing tools is a different story. And the test isn’t finished yet. The current run is approaching 400K tokens, with the expectation that it can keep going. This is also another interesting example of why Apple Silicon keeps showing up in local LLM experiments. The M5 Max’s unified memory gives a large model and its growing KV cache access to one shared memory pool. With MLX-Serve continuing to improve, these machines are becoming surprisingly capable long-context inference boxes. The bigger takeaway: Context length is becoming a workload, not just a model specification. Running a model at 256K is one thing. Keeping an agent alive at 300K+ while it reasons and uses tools is much more interesting. And Qwen3.8-Flash-Next is showing that this can be pushed surprisingly far on a single 128GB Mac. 364.7K and counting. Next stop: 400K.

Qwen3.8-Flash-Next is still going strong at 364.7K tokens of context on an M5 Max. And this isn’t just a static long-context test. The model was reasoning about how to speed up its own workflow while using tools, and the tool calls kept working without misses. Setup: • Qwen3.8-Flash-Next • M5 Max • 128GB unified memory • MLX-Serve PR #363 • OpenCode 2 • 364.7K context The interesting part isn’t simply getting hundreds of thousands of tokens into memory. It’s what happens once the context gets this large. Long-context inference usually comes with a painful tradeoff. As the KV cache grows, memory pressure increases and generation can slow down. But this setup is still pushing through 364K tokens while maintaining a usable agent workflow. The model can reason, call tools, inspect results, continue working, and keep the session moving. And the tool calls reportedly haven’t missed so far. That’s important for agentic coding. A huge context window is only useful if the model can actually operate reliably inside it. A 400K-token context that constantly breaks tool calls isn’t very useful. A 364K session that can keep reasoning and executing tools is a different story. And the test isn’t finished yet. The current run is approaching 400K tokens, with the expectation that it can keep going. This is also another interesting example of why Apple Silicon keeps showing up in local LLM experiments. The M5 Max’s unified memory gives a large model and its growing KV cache access to one shared memory pool. With MLX-Serve continuing to improve, these machines are becoming surprisingly capable long-context inference boxes. The bigger takeaway: Context length is becoming a workload, not just a model specification. Running a model at 256K is one thing. Keeping an agent alive at 300K+ while it reasons and uses tools is much more interesting. And Qwen3.8-Flash-Next is showing that this can be pushed surprisingly far on a single 128GB Mac. 364.7K and counting. Next stop: 400K.

FHILY👑

39,982 Aufrufe • vor 10 Tagen

Cerebras inference is very fast. So fast that it changes how we think about configuring our LLMs for voice agent use cases. Kimi K2.6 is a 1T parameter reasoning model that Cerebras serves at 650 - 1,000 tokens per second (end-to-end throughput), with time to first token metrics as low as 150ms (latency). These numbers are two to three times faster than other similarly capable models. The biggest lever we get from this kind of speed is that we can use the model in reasoning mode, and still have excellent "time to first non-thinking token." This solves a big pain point we have in 2026 for voice agent use cases. Almost all recent innovation in post-training has focused on making models good at reasoning ("test time compute"). This is great, but it makes the user-facing model latency much, much slower. Which is a problem for conversational voice agents. We can run Kimi K2.6 with reasoning turned on, and get responses faster than other models produce with reasoning disabled. On my 30-turn voice agent benchmark, Kimi K2.6 with reasoning enabled ties GPT 5.1 and Haiku 4.5 with reasoning disabled, and is still about 200ms seconds faster! On my primary task agent benchmark, Kimi K2.6 is now the #2 model. It ranks just behind Gemini 3.5 Flash in "high" reasoning mode, and tied with GLM 5, Sonnet 4.6, and GPT 5.4 with reasoning set to "low." But Kimi K2.6 completes each turn in the agent loop in under 500ms. The other four models are all at least 3x slower. (Models only qualify for this benchmark if they can complete task turns at a P50 <4s.) A couple of other things that this speed buys us, for production voice agents: - Tool calls happen fast enough that we don't have to work around tool call latency in our pipeline design. - We can prompt the model to output structured data at the beginning of a response, followed by plain text for voice generation. This opens up possibilities like asking the model to do complex classification/generation tasks that influence the rest of the pipeline. For example, the model could create a detailed style prompt for a steerable TTS model, for each individual conversation turn. And, of course, you can use Kimi K2.6 with reasoning turned off. Cerebras calls this "instant" mode. Here's a video of a Cerebras Kimi K2.6 voice agent with voice-to-voice response time, measured at the client, under 500ms. This is the true response latency as perceived by the user, including all network and audio codec overhead, transcription and turn detection, Kimi K2.6 token generation, and voice generation. 500ms is, effectively, instant. So the Cerebras naming for this mode is a propos. :-)

Cerebras inference is very fast. So fast that it changes how we think about configuring our LLMs for voice agent use cases. Kimi K2.6 is a 1T parameter reasoning model that Cerebras serves at 650 - 1,000 tokens per second (end-to-end throughput), with time to first token metrics as low as 150ms (latency). These numbers are two to three times faster than other similarly capable models. The biggest lever we get from this kind of speed is that we can use the model in reasoning mode, and still have excellent "time to first non-thinking token." This solves a big pain point we have in 2026 for voice agent use cases. Almost all recent innovation in post-training has focused on making models good at reasoning ("test time compute"). This is great, but it makes the user-facing model latency much, much slower. Which is a problem for conversational voice agents. We can run Kimi K2.6 with reasoning turned on, and get responses faster than other models produce with reasoning disabled. On my 30-turn voice agent benchmark, Kimi K2.6 with reasoning enabled ties GPT 5.1 and Haiku 4.5 with reasoning disabled, and is still about 200ms seconds faster! On my primary task agent benchmark, Kimi K2.6 is now the #2 model. It ranks just behind Gemini 3.5 Flash in "high" reasoning mode, and tied with GLM 5, Sonnet 4.6, and GPT 5.4 with reasoning set to "low." But Kimi K2.6 completes each turn in the agent loop in under 500ms. The other four models are all at least 3x slower. (Models only qualify for this benchmark if they can complete task turns at a P50 <4s.) A couple of other things that this speed buys us, for production voice agents: - Tool calls happen fast enough that we don't have to work around tool call latency in our pipeline design. - We can prompt the model to output structured data at the beginning of a response, followed by plain text for voice generation. This opens up possibilities like asking the model to do complex classification/generation tasks that influence the rest of the pipeline. For example, the model could create a detailed style prompt for a steerable TTS model, for each individual conversation turn. And, of course, you can use Kimi K2.6 with reasoning turned off. Cerebras calls this "instant" mode. Here's a video of a Cerebras Kimi K2.6 voice agent with voice-to-voice response time, measured at the client, under 500ms. This is the true response latency as perceived by the user, including all network and audio codec overhead, transcription and turn detection, Kimi K2.6 token generation, and voice generation. 500ms is, effectively, instant. So the Cerebras naming for this mode is a propos. :-)

kwindla

40,593 Aufrufe • vor 3 Monaten

ChatGPT o1 is the most "intelligent" AI model and it's not even close! Full o1 generates thinking steps ~50 faster than preview. It's more accurate, reliable, and got better on harder tasks that require advanced reasoning and knowledge. I ran a few tests on it already. Here are my observations: Full video with examples & explanations: Strengths - impressive at math, code, and knowledge-intensive tasks. Weakness - it only failed on a cross-word puzzle but I think it might be solvable when a web search becomes available. In the end, while very efficient with complex knowledge use, it's still constrained by data it's trained on. Speed - the thinking steps are generated a lot faster! Not a fair comparison with the open alternatives but I think this improves the overall user experience. "Knowledgeable and highly intelligent" - as mentioned in the demo by OpenAI researchers, o1 is great at dealing with ambiguity and filling in knowledge gaps. I was impressed by how it implemented an agentic solution (with lots of details) from a basic diagram of architecture (with minimal details). Check out the sample video. Better Task Coverage - Due to the speed and the ability to make sense of instructions and intent (i.e., know when to response fast and when to "think" deeply) much better, it feels like it might be more useful for a broader range of tasks. Image understanding - the image understanding capability is mysterious (often leads to faster responses but no thinking) but impressive. More experiments and notes soon. Stay tuned!

ChatGPT o1 is the most "intelligent" AI model and it's not even close! Full o1 generates thinking steps ~50 faster than preview. It's more accurate, reliable, and got better on harder tasks that require advanced reasoning and knowledge. I ran a few tests on it already. Here are my observations: Full video with examples & explanations: Strengths - impressive at math, code, and knowledge-intensive tasks. Weakness - it only failed on a cross-word puzzle but I think it might be solvable when a web search becomes available. In the end, while very efficient with complex knowledge use, it's still constrained by data it's trained on. Speed - the thinking steps are generated a lot faster! Not a fair comparison with the open alternatives but I think this improves the overall user experience. "Knowledgeable and highly intelligent" - as mentioned in the demo by OpenAI researchers, o1 is great at dealing with ambiguity and filling in knowledge gaps. I was impressed by how it implemented an agentic solution (with lots of details) from a basic diagram of architecture (with minimal details). Check out the sample video. Better Task Coverage - Due to the speed and the ability to make sense of instructions and intent (i.e., know when to response fast and when to "think" deeply) much better, it feels like it might be more useful for a broader range of tasks. Image understanding - the image understanding capability is mysterious (often leads to faster responses but no thinking) but impressive. More experiments and notes soon. Stay tuned!

elvis

102,643 Aufrufe • vor 1 Jahr