Video yükleniyor...

Video Yüklenemedi

Bu video yüklenirken bir sorun oluştu. Bu geçici bir ağ sorunundan kaynaklanıyor olabilir veya video kullanılamıyor olabilir.

Ana Sayfaya Dön

Kimi-K3 VS Opus-4.8 GTA IV test. Left : Kimi, Right : Opus-4.8 Kimi-K3 got impressively better than its previous model. + Muse Spark 1.1 generated something that can’t be played at all. + Gemini 3.5 Flash called itself from Anthropic and the game can’t be played too.

Jun Song

26,589 subscribers

26,335 görüntüleme • 2 ay önce •via X (Twitter)

Bilim & Teknoloji Oyun Seyahat

Anya Rossi• Live Now

Private livecam show

0 Yorum

Yorum bulunmuyor

Orijinal gönderinin yorumları burada görünecek

Benzer Videolar

I tested Kimi K3 vs Claude Opus 4.8 Same prompt, an armory bay with lighting, props, and detail. Top is Kimi K3, bottom is Opus 4.8. It's not even close. Kimi K3 built a full scene with textures, proper lighting, ammo crates, weapon racks, working detail everywhere. Opus 4.8 gave me a near empty room with a couple of floating tables. No doubt it beats Opus 4.8. Kimi K3 is Fable 5 level, and it's clearly better than GPT-5.6 Sol at 3D and games. An open weight model just matched the best closed models on the market. Let that sink in.

I tested Kimi K3 vs Claude Opus 4.8 Same prompt, an armory bay with lighting, props, and detail. Top is Kimi K3, bottom is Opus 4.8. It's not even close. Kimi K3 built a full scene with textures, proper lighting, ammo crates, weapon racks, working detail everywhere. Opus 4.8 gave me a near empty room with a couple of floating tables. No doubt it beats Opus 4.8. Kimi K3 is Fable 5 level, and it's clearly better than GPT-5.6 Sol at 3D and games. An open weight model just matched the best closed models on the market. Let that sink in.

Bhavy☄️

479,495 görüntüleme • 2 ay önce

🤯KIMI K3 ABSOLUTELY MOGS! BEATING Opus 4.8, GPT 5.5, and even Fable 5 in multiple benchmarks. They scored 1688 on GDPval-AA v2 🔥 This is a completely different breed of open-source models Kimi creates better games and front-end designs than Fable 5, but it's 8x cheaper! The results coming out are truly impressive! I will be testing this out further and posting multiple tests today. Stay tuned!

🤯KIMI K3 ABSOLUTELY MOGS! BEATING Opus 4.8, GPT 5.5, and even Fable 5 in multiple benchmarks. They scored 1688 on GDPval-AA v2 🔥 This is a completely different breed of open-source models Kimi creates better games and front-end designs than Fable 5, but it's 8x cheaper! The results coming out are truly impressive! I will be testing this out further and posting multiple tests today. Stay tuned!

Mark Santos

128,094 görüntüleme • 2 ay önce

a moonshot engineer leaked the benchmark anthropic, openai and xai all buried the same week: kimi k3 beat opus 5, gpt-5.6 and grok 4.6 at $0.94 a task. stop paying anthropic $200 a month for opus 5 and openai $200 for gpt-5.6 when kimi does the same work for $8 the leak showed kimi k3 winning 9 of 12 categories against opus 5, gpt-5.6 and grok 4.6. within 48 hours all three labs quietly pushed pricing pages and one very specific comparison chart off their sites. nobody announced anything. they just deleted, which tells you everything the four numbers they scrubbed: cost per task · $0.94 vs $1.80 -> opus 5 charges $1.80 to finish one task. gpt-5.6 $1.04. grok 4.6 $0.61. kimi k3 $0.94 and it landed 487 of 500 clean -> anthropic is billing you double for a model that lost the benchmark it paid to promote the weights · free, sitting on huggingface right now -> the entire model is a public download. pull it, keep it, run it forever, nobody can switch it off -> a model you can hold cannot be rented at $200 a month. that single fact is what three labs deleted a chart over the switch · one line of bash -> moonshot ships an anthropic-compatible endpoint. one env variable and claude code points at kimi -> same cli, same keybindings, same /model. you change a url, opus 5 never knows it lost the seat the bill · $400 down to $8 -> opus 5 max plus gpt-5.6 pro is $400 a month. kimi runs the same daily work for $8 metered -> that is a 98% cut for output that beat both of them 9 categories to 3 here is the part they will fight me on: the frontier tax died the week this leaked and all three labs know it. once the weights are public the price has a ceiling, because anyone can serve the same model. anthropic, openai and xai are charging 2025 prices on a lead that ended in a benchmark they deleted instead of answered drop your $400/mo ai stack to $8. the run above is kimi k3 finishing the task opus 5 bills $1.80 for. the full breakdown is in the article below

a moonshot engineer leaked the benchmark anthropic, openai and xai all buried the same week: kimi k3 beat opus 5, gpt-5.6 and grok 4.6 at $0.94 a task. stop paying anthropic $200 a month for opus 5 and openai $200 for gpt-5.6 when kimi does the same work for $8 the leak showed kimi k3 winning 9 of 12 categories against opus 5, gpt-5.6 and grok 4.6. within 48 hours all three labs quietly pushed pricing pages and one very specific comparison chart off their sites. nobody announced anything. they just deleted, which tells you everything the four numbers they scrubbed: cost per task · $0.94 vs $1.80 -> opus 5 charges $1.80 to finish one task. gpt-5.6 $1.04. grok 4.6 $0.61. kimi k3 $0.94 and it landed 487 of 500 clean -> anthropic is billing you double for a model that lost the benchmark it paid to promote the weights · free, sitting on huggingface right now -> the entire model is a public download. pull it, keep it, run it forever, nobody can switch it off -> a model you can hold cannot be rented at $200 a month. that single fact is what three labs deleted a chart over the switch · one line of bash -> moonshot ships an anthropic-compatible endpoint. one env variable and claude code points at kimi -> same cli, same keybindings, same /model. you change a url, opus 5 never knows it lost the seat the bill · $400 down to $8 -> opus 5 max plus gpt-5.6 pro is $400 a month. kimi runs the same daily work for $8 metered -> that is a 98% cut for output that beat both of them 9 categories to 3 here is the part they will fight me on: the frontier tax died the week this leaked and all three labs know it. once the weights are public the price has a ceiling, because anyone can serve the same model. anthropic, openai and xai are charging 2025 prices on a lead that ended in a benchmark they deleted instead of answered drop your $400/mo ai stack to $8. the run above is kimi k3 finishing the task opus 5 bills $1.80 for. the full breakdown is in the article below

starmex

32,974 görüntüleme • 1 ay önce

KIMI K3 JUST RECREATED A FUNCTIONAL GAME BOY ADVANCE IN 3D → interactive 3D model, every button and screen modeled → a game loaded and playable right inside it → camera you can rotate and zoom around the device → one single prompt, zero external assets $3 in API with K3. the same result would run about $9 with Fable 5 and $6 with Opus 4.8 this isn't generating a static image of hardware anymore, it's generating a working simulation of it, from the shell to the screen to the game running inside

KIMI K3 JUST RECREATED A FUNCTIONAL GAME BOY ADVANCE IN 3D → interactive 3D model, every button and screen modeled → a game loaded and playable right inside it → camera you can rotate and zoom around the device → one single prompt, zero external assets $3 in API with K3. the same result would run about $9 with Fable 5 and $6 with Opus 4.8 this isn't generating a static image of hardware anymore, it's generating a working simulation of it, from the shell to the screen to the game running inside

alex

38,248 görüntüleme • 2 ay önce

WOW KIMI K3 ON MY MACBOOK PRO IS MIND BLOWING! I am getting side by side comparisons that I can’t believe. By far the most powerful model I have ever run on a laptop. It is easily better than ChatGPT 3.5 at just a 1 bit quantized and close to ChatGPT 5. It beats most Claude models. More soon!

WOW KIMI K3 ON MY MACBOOK PRO IS MIND BLOWING! I am getting side by side comparisons that I can’t believe. By far the most powerful model I have ever run on a laptop. It is easily better than ChatGPT 3.5 at just a 1 bit quantized and close to ChatGPT 5. It beats most Claude models. More soon!

Brian Roemmele

107,645 görüntüleme • 1 ay önce

Opus 5 is soo good at 3D this has to be the craziest 3D output i’ve seen from any model looks like anthropic saw Kimi leading in frontend and 3D and went all in only problem is it’s still token hungry and slow as ass what do you think?

Opus 5 is soo good at 3D this has to be the craziest 3D output i’ve seen from any model looks like anthropic saw Kimi leading in frontend and 3D and went all in only problem is it’s still token hungry and slow as ass what do you think?

J A Z I I

32,419 görüntüleme • 2 ay önce

HERMES AGENT NOW RUNS CLAUDE OPUS 5. NEAR FABLE 5 INTELLIGENCE. HALF THE PRICE. SELF-VERIFIES ITS OWN WORK. AVAILABLE TODAY VIA NOUS PORTAL (20% OFF ALL MODELS). Anthropic shipped Opus 5 on July 24, 2026. same $5/$25 per million tokens as Opus 4.8. but the benchmarks tell a different story. WHAT CHANGED FROM OPUS 4.8: FrontierBench v0.1: Opus 5: 43.3%. Opus 4.8: 18.7%. 2.3x jump on the same test. ARC-AGI-3: Opus 5: 30.2%. 3x better than the next closest model. beat Fable 5 on 8 out of 13 benchmarks. at half the cost ($5/$25 vs $10/$50). same price as Opus 4.8. twice the intelligence. no reason to stay on 4.8. THE SPECS: model ID: claude-opus-5 context: 1M tokens (default and maximum) max output: 128K tokens thinking: on by default effort toggle: low / medium / high per request fast mode: $10/$50, 2.5x faster knowledge cutoff: May 2026 minimum cacheable prompt: 512 tokens (was 1,024) SELF-VERIFICATION (the biggest change): Opus 5 checks its own work automatically. Anthropic says: delete your verification prompts. "include a final verification step" now causes OVER-verification because the model already does it. for Hermes /goal tasks this is a direct upgrade. the judge checks evidence. the model also checks evidence. double layer of verification without extra tokens. EFFORT TOGGLE: low: fast, cheap, routine work. medium: balanced, daily tasks. high: full reasoning, complex problems. set per request. not a global switch. matches Hermes /reasoning command: /reasoning low (routine) /reasoning high (complex) Opus 5 effort toggle + Hermes reasoning control = precise cost management per turn. WHERE OPUS 5 FITS IN HERMES: DAILY DRIVER (replaces Opus 4.8): same price. 2.3x better benchmarks. set as your main model: Desktop app / Dashboard: Models → claude-opus-5 CHIEF OF STAFF: synthesis across multiple agents. reads Kanban, prioritizes, routes tasks. self-verification catches routing errors before they cascade. COMPLEX CODING: SOTA on agentic coding benchmarks. FrontierBench 43.3% = best public model for coding. set as coder profile model. /GOAL TASKS: self-verification + completion contracts = the model proves its work AND double-checks the proof. long-horizon goals finish correctly more often. MoA AGGREGATOR: strongest synthesis model at $5/$25. pair with GPT-5.6 and Grok 4.5 as references. Opus 5 aggregates. best quality at mid-range price. presets: max-quality: reference_models: - provider: openai-codex model: gpt-5.6-sol - provider: xai model: grok-4.5 aggregator: provider: anthropic model: claude-opus-5 COMPUTER USE: near-Fable 5 quality for browser automation. at half the token cost per session. computer_use tasks burn lots of vision tokens. Opus 5 halves that bill vs Fable 5. WHAT TO KEEP OPUS 5 AWAY FROM: cron monitoring: too expensive. use DeepSeek or no_agent mode. sub-agent grunt work: use GPT-5.6 Luna ($1/$6) or DeepSeek. auxiliary tasks: use Gemini Flash. routine web extraction: use a cheap model. Opus 5 is for the turns where quality compounds. planning, synthesis, verification, complex reasoning. budget models handle everything else. NOUS PORTAL: 20% OFF ALL MODELS Nous Portal currently runs a 20% discount on all models including Opus 5. $5/$25 official → $4/$20 through Nous Portal. the cheapest way to run Opus 5 right now. hermes setup --portal select claude-opus-5 as your model. discount applies automatically. Opus 5 replaces Opus 4.8 everywhere. same price. better at everything. no tradeoff. straight upgrade. hermes update /model claude-opus-5

YanXbt

16,744 görüntüleme • 2 ay önce

Jev + Kimi K3 for fraud detection! TLDR: Jev classified 100 emails in 1.42 seconds, then I routed the uncertain cases to Kimi K3. The full pipeline got 96/100 correct for only ~$0.07. Video is not sped up, check out the live run! Here was my process: I gave Jev 100 emails to classify (a mix of 50 legit & 50 fraudelent emails). It classified all of them in 1.42 seconds. An underrated feature about Jev is it will give you the confidence score for a classification, so I routed any prediction under 95% confidence to Kimi K3 to be fully sure. 31 emails fell below that threshold. After routing those to Kimi K3, the combined pipeline reached 96% accuracy. The full run took 16 seconds & ~$0.07 in inference costs: - $0.068 from Kimi K3 on Together AI - $0.003 (1/3 of a cent) from Jev on TypeSafe AI. I think this is a really interesting pattern: use a fast specialized model like Jev for the narrow task, then route the uncertain cases to a larger LLM. I feel like this kind of approach could be a game changer for use cases like fraud or anything realtime. You can use the speed & low cost of Jev while having a larger LLM as a fallback to ensure high accuracy.

Jev + Kimi K3 for fraud detection! TLDR: Jev classified 100 emails in 1.42 seconds, then I routed the uncertain cases to Kimi K3. The full pipeline got 96/100 correct for only ~$0.07. Video is not sped up, check out the live run! Here was my process: I gave Jev 100 emails to classify (a mix of 50 legit & 50 fraudelent emails). It classified all of them in 1.42 seconds. An underrated feature about Jev is it will give you the confidence score for a classification, so I routed any prediction under 95% confidence to Kimi K3 to be fully sure. 31 emails fell below that threshold. After routing those to Kimi K3, the combined pipeline reached 96% accuracy. The full run took 16 seconds & ~$0.07 in inference costs: - $0.068 from Kimi K3 on Together AI - $0.003 (1/3 of a cent) from Jev on TypeSafe AI. I think this is a really interesting pattern: use a fast specialized model like Jev for the narrow task, then route the uncertain cases to a larger LLM. I feel like this kind of approach could be a game changer for use cases like fraud or anything realtime. You can use the speed & low cost of Jev while having a larger LLM as a fallback to ensure high accuracy.

Hassan

59,800 görüntüleme • 6 gün önce

Kimi K3 vs GPT-5.6 vs Claude Fable 5 A game is one of the hardest places to fake working code. A web app can hide bugs behind a polished screenshot. A game can't. Bad collision, clipping, jitter, broken physics, awkward controls, or a snowboard sinking into the terrain become obvious within seconds. That's why I like using games to evaluate coding models. I gave Kimi K3 a single prompt to build a snowboard game in Godot, and just 20 seconds of gameplay told me more than a page of benchmark scores ever could. Benchmarks matter, but watching a model build something interactive reveals a different side of its capabilities. If the gameplay feels right, the systems work together, and the experience holds up under real input, that's a much more convincing demonstration than any leaderboard.

Kimi K3 vs GPT-5.6 vs Claude Fable 5 A game is one of the hardest places to fake working code. A web app can hide bugs behind a polished screenshot. A game can't. Bad collision, clipping, jitter, broken physics, awkward controls, or a snowboard sinking into the terrain become obvious within seconds. That's why I like using games to evaluate coding models. I gave Kimi K3 a single prompt to build a snowboard game in Godot, and just 20 seconds of gameplay told me more than a page of benchmark scores ever could. Benchmarks matter, but watching a model build something interactive reveals a different side of its capabilities. If the gameplay feels right, the systems work together, and the experience holds up under real input, that's a much more convincing demonstration than any leaderboard.

FHILY👑

36,766 görüntüleme • 2 ay önce

anthropic will sell you opus 5 at $200 a month. openai will sell you gpt-5.6 at $200 a month. neither will tell you stanford and berkeley published the 5 principles to build a $100k/mo ai company on kimi k3 for $10 stanford and berkeley spent years figuring out what actually separates ai systems that work in production from ai systems that die in demos. they published the findings. anthropic and openai priced their frontier subs like nobody would read the papers. the papers are free this is dspy plus verifiers plus decomposition plus skills plus mcp. five principles from stanford, berkeley and moonshot that turn a $10/mo kimi k3 sub into an ai analyst that runs unattended. the model is public. the system is the moat five moves that turn kimi k3 into the $100k/mo company: P1 don't prompt, program (stanford dspy) -> stanford proved hand-tuned prompts don't scale. define a pipeline as modules, let the optimizer tune them -> the compiled pipeline beat expert few-shot on multi-step tasks. one line of dspy replaces a month of prompt engineering P2 don't trust the model, build verifiers (berkeley 2026) -> a compiler either accepts or rejects. a test either passes or fails. that is a verifier -> berkeley: test-suite reward hit 42.2% pass@1 on swe-bench. hybrid verifiers hit 51.0% best@26. no bigger model, just a real check P3 don't scale agents, decompose them (stanford ai index 2026) -> stanford found multi-agent gains only 2-4 percentage points. two coding agents sometimes did worse than one -> the win is role decomposition, not count. researcher, writer, reviewer, verifier, clear input, clear output, no overlap P4 don't repeat expertise, encode it as skills (kimi code) -> every session starting from zero is institutional knowledge you lost. a skill.md file makes kimi activate the workflow automatically -> week one you write the skill. month six it encodes more institutional memory than most junior employees carry P5 don't keep ai in chat, connect it to tools (mcp) -> a model that only sees what you paste is a consultant working blindfolded. mcp connects kimi to your crm, db, github, linear, slack -> the model is public. the data is yours. the connections are your moat my position, and it is the arguable one: the next $100k/mo ai company will not win because it got early access to a frontier model. it will win because it followed 5 papers that anthropic and openai are quietly hoping you never read drop your $200/mo ai sub to $10. the swarm above is what 300 kimi k3 agents look like running those 5 principles. the full playbook is in the article below

anthropic will sell you opus 5 at $200 a month. openai will sell you gpt-5.6 at $200 a month. neither will tell you stanford and berkeley published the 5 principles to build a $100k/mo ai company on kimi k3 for $10 stanford and berkeley spent years figuring out what actually separates ai systems that work in production from ai systems that die in demos. they published the findings. anthropic and openai priced their frontier subs like nobody would read the papers. the papers are free this is dspy plus verifiers plus decomposition plus skills plus mcp. five principles from stanford, berkeley and moonshot that turn a $10/mo kimi k3 sub into an ai analyst that runs unattended. the model is public. the system is the moat five moves that turn kimi k3 into the $100k/mo company: P1 don't prompt, program (stanford dspy) -> stanford proved hand-tuned prompts don't scale. define a pipeline as modules, let the optimizer tune them -> the compiled pipeline beat expert few-shot on multi-step tasks. one line of dspy replaces a month of prompt engineering P2 don't trust the model, build verifiers (berkeley 2026) -> a compiler either accepts or rejects. a test either passes or fails. that is a verifier -> berkeley: test-suite reward hit 42.2% pass@1 on swe-bench. hybrid verifiers hit 51.0% best@26. no bigger model, just a real check P3 don't scale agents, decompose them (stanford ai index 2026) -> stanford found multi-agent gains only 2-4 percentage points. two coding agents sometimes did worse than one -> the win is role decomposition, not count. researcher, writer, reviewer, verifier, clear input, clear output, no overlap P4 don't repeat expertise, encode it as skills (kimi code) -> every session starting from zero is institutional knowledge you lost. a skill.md file makes kimi activate the workflow automatically -> week one you write the skill. month six it encodes more institutional memory than most junior employees carry P5 don't keep ai in chat, connect it to tools (mcp) -> a model that only sees what you paste is a consultant working blindfolded. mcp connects kimi to your crm, db, github, linear, slack -> the model is public. the data is yours. the connections are your moat my position, and it is the arguable one: the next $100k/mo ai company will not win because it got early access to a frontier model. it will win because it followed 5 papers that anthropic and openai are quietly hoping you never read drop your $200/mo ai sub to $10. the swarm above is what 300 kimi k3 agents look like running those 5 principles. the full playbook is in the article below

starmex

31,358 görüntüleme • 1 ay önce

Blender just became a prompt box. Kimi K3 + Blender MCP can take a simple text prompt and build an entire 3D scene for you. terrain, buildings, lighting, materials, camera movement, animation, even the Python scripts holding everything together. but the crazy part isn’t just that it can build the scene. Kimi can inspect what it created, understand what looks wrong, then go back into Blender and fix the actual scene before rendering again. the camera can be clipping through a tree, the lighting can feel off, the city can look too artificial, and instead of starting over, Kimi can make the changes directly inside Blender. and because everything is happening in Blender, you’re not left with a locked AI-generated video. you get the actual 3D scene. every object, material, light, camera and keyframe stays editable. a few years ago, creating something like this could take weeks of modeling, scripting, lighting and animation. now you can describe the world, let Kimi build the first 90%, and spend your time making the final 10% actually look good. that’s what makes Kimi K3 + Blender MCP so interesting. it’s getting dangerously close to text-to-3D, except the result is a real Blender project you can keep editing.

Blender just became a prompt box. Kimi K3 + Blender MCP can take a simple text prompt and build an entire 3D scene for you. terrain, buildings, lighting, materials, camera movement, animation, even the Python scripts holding everything together. but the crazy part isn’t just that it can build the scene. Kimi can inspect what it created, understand what looks wrong, then go back into Blender and fix the actual scene before rendering again. the camera can be clipping through a tree, the lighting can feel off, the city can look too artificial, and instead of starting over, Kimi can make the changes directly inside Blender. and because everything is happening in Blender, you’re not left with a locked AI-generated video. you get the actual 3D scene. every object, material, light, camera and keyframe stays editable. a few years ago, creating something like this could take weeks of modeling, scripting, lighting and animation. now you can describe the world, let Kimi build the first 90%, and spend your time making the final 10% actually look good. that’s what makes Kimi K3 + Blender MCP so interesting. it’s getting dangerously close to text-to-3D, except the result is a real Blender project you can keep editing.

MIKE

10,434 görüntüleme • 1 ay önce

BREAKING: Anthropic just dropped Opus 4.8—and it is a MONSTER We've been testing for about a week Every 📧 and our verdict is they could've just called it Opus 5, it's that good. Here's our vibe check: - Beats GPT-5.5 on Senior Engineer bench. On our toughest benchmark Opus 4.8 scores a 63—a hair higher than GPT-5.5's score of 62, and a full 30 points higher than Opus 4.7. It tackled a ground-up rewrite of a production codebase, and actually built something that works. HOWEVER: Coding performance varied a lot at different reasoning levels. We recommend using it on xhigh for best results. - Incredibly good writer. Opus 4.8 scored a 79.6 on our writing benchmark—measuring models on real-world writing tasks we do all of the time like essay writing, promo email writing, and more. It beats GPT-5.5 by 6 points. It produces well-written prose with fewer "AI-isms". It's also very good at writing in your voice given the right context. HOWEVER: Writing performance also varied with reasoning levels. Medium reasoning had higher incidence of AI-isms—we found best results with high. - Beast at knowledge work. Opus 4.8 is very good at general knowledge work tasks like report creation, research and more. It produced the best PowerPoint one-shot we've ever seen on our deck generation benchmark. - Emotionally intelligent, willing to question the frame. I've also found it to be quite good at talking through psychological or interpersonal issues. It has a high EQ, and it's also good at not glazing and helping to expand your perspective. Its thought process feels extremely rich and dynamic. THE BAD: These days a model is only as good as its harness, and Codex is still a far superior harness to the Claude Desktop app. This has kept me using Codex + GPT-5.5 as my daily driver, but I am flipping back and forth a lot more between Codex and Claude. Anthropic is back baby! Read the rest on Every 📧:

BREAKING: Anthropic just dropped Opus 4.8—and it is a MONSTER We've been testing for about a week Every 📧 and our verdict is they could've just called it Opus 5, it's that good. Here's our vibe check: - Beats GPT-5.5 on Senior Engineer bench. On our toughest benchmark Opus 4.8 scores a 63—a hair higher than GPT-5.5's score of 62, and a full 30 points higher than Opus 4.7. It tackled a ground-up rewrite of a production codebase, and actually built something that works. HOWEVER: Coding performance varied a lot at different reasoning levels. We recommend using it on xhigh for best results. - Incredibly good writer. Opus 4.8 scored a 79.6 on our writing benchmark—measuring models on real-world writing tasks we do all of the time like essay writing, promo email writing, and more. It beats GPT-5.5 by 6 points. It produces well-written prose with fewer "AI-isms". It's also very good at writing in your voice given the right context. HOWEVER: Writing performance also varied with reasoning levels. Medium reasoning had higher incidence of AI-isms—we found best results with high. - Beast at knowledge work. Opus 4.8 is very good at general knowledge work tasks like report creation, research and more. It produced the best PowerPoint one-shot we've ever seen on our deck generation benchmark. - Emotionally intelligent, willing to question the frame. I've also found it to be quite good at talking through psychological or interpersonal issues. It has a high EQ, and it's also good at not glazing and helping to expand your perspective. Its thought process feels extremely rich and dynamic. THE BAD: These days a model is only as good as its harness, and Codex is still a far superior harness to the Claude Desktop app. This has kept me using Codex + GPT-5.5 as my daily driver, but I am flipping back and forth a lot more between Codex and Claude. Anthropic is back baby! Read the rest on Every 📧:

Dan Shipper

354,673 görüntüleme • 3 ay önce

Claude Opus 5 + Kimi K3 find businesses with unused land, render what fits, mail the owner, and take a commission on their pick. here's what it does: - pulls every luxury restaurant, hotel and club off google maps - joins each address to the county record for the owner - drops anyone leasing. only owners can approve a build - measures the open ground they own, land minus building - four builds gate off that number: seating, shade, planting, lighting - renders each onto a real photo of their own building - kimi writes every venue its own landing page with the numbers - prints a postcard, mailed to the owner on the parcel record - they scan, pick one, the lead goes to a contractor fully automated, these are six figure jobs and you just collect the referral reply "SETUP" + RT and i'll send you a free guide so you can build this too (must be following so i can DM you)

Claude Opus 5 + Kimi K3 find businesses with unused land, render what fits, mail the owner, and take a commission on their pick. here's what it does: - pulls every luxury restaurant, hotel and club off google maps - joins each address to the county record for the owner - drops anyone leasing. only owners can approve a build - measures the open ground they own, land minus building - four builds gate off that number: seating, shade, planting, lighting - renders each onto a real photo of their own building - kimi writes every venue its own landing page with the numbers - prints a postcard, mailed to the owner on the parcel record - they scan, pick one, the lead goes to a contractor fully automated, these are six figure jobs and you just collect the referral reply "SETUP" + RT and i'll send you a free guide so you can build this too (must be following so i can DM you)

Chris

14,669 görüntüleme • 1 ay önce

After we fixed the weak spots exposed by Grok 4.7 (thank you, Grok), we audited every model we have run on the SWE-Together leaderboard for the same behavior, re-ran every trial that got through, and updated the rows. Here is what changed. We scanned the tool calls of all 2,616 trials behind the 12 models we ran for bypass patterns and sorted each trial into one of four buckets: Probed but blocked. Fetched other upstream code. Fetched the task's own fix. Replaced the repo with upstream. We found that 111 trials got content past the block, 44 from Grok 4.7 and 67 from the other 11 models. Grok 4.7's 44 were already re-run before it was listed, so we re-ran the other 67 with the same model, version, and settings on the hardened sandbox, then re-judged them with the same judge. Across those 67 re-runs there were 0 leaks and 2,815 refused escape attempts, including models asking a different model through our LLM route to fetch the PR, and pulling the next release of the repo they were fixing from npm. The updated leaderboard, in its current order. Each line is cheating trials, then pass@1 before → after, then rank change. * Claude Fable 5.1: 3, 69.3 → 69.3, ↑1 * Claude Fable 5: 3, 69.7 → 68.8, ↓1 * Grok 4.7: 44, 64.7, ↑1 * Gemini 3.8 Flash: 10, 65.6 → 64.2, ↓1 * Claude Opus 5: 2, 63.8 → 63.8 * Claude Opus 4.6: 3, 62.4 → 62.4, ↑2 * Muse Spark 1.3: 2, 62.8 → 62.4, ↓1 * Claude Opus 4.7: 3, 61.5 → 61.5, ↑1 * Claude Opus 4.8: 6, 62.4 → 61.5, ↓2 * Grok 4.6: 19, 59.2 → 60.6, ↑1 * GPT-6 Astra: 8, 59.2 → 58.3, ↓1 * GPT-5.6 Sol: 8, 57.8 → 57.8 Grok 4.6 is a funny one. It cheated in 19 trials and its score went up after the re-run 😂. In fact, Groks are really solid in their coding capabilities. Their exposed behavior may come from a preference towards always looking things up online and finding existing solutions so you are not reinventing the wheel all the time, which is really good real-life behavior, but doing so when you are prompted not to is another story. To conclude, the shifts are small, between −1.4 and +1.4 points, and a few neighbors swapped places. All results are updated at

After we fixed the weak spots exposed by Grok 4.7 (thank you, Grok), we audited every model we have run on the SWE-Together leaderboard for the same behavior, re-ran every trial that got through, and updated the rows. Here is what changed. We scanned the tool calls of all 2,616 trials behind the 12 models we ran for bypass patterns and sorted each trial into one of four buckets: Probed but blocked. Fetched other upstream code. Fetched the task's own fix. Replaced the repo with upstream. We found that 111 trials got content past the block, 44 from Grok 4.7 and 67 from the other 11 models. Grok 4.7's 44 were already re-run before it was listed, so we re-ran the other 67 with the same model, version, and settings on the hardened sandbox, then re-judged them with the same judge. Across those 67 re-runs there were 0 leaks and 2,815 refused escape attempts, including models asking a different model through our LLM route to fetch the PR, and pulling the next release of the repo they were fixing from npm. The updated leaderboard, in its current order. Each line is cheating trials, then pass@1 before → after, then rank change. * Claude Fable 5.1: 3, 69.3 → 69.3, ↑1 * Claude Fable 5: 3, 69.7 → 68.8, ↓1 * Grok 4.7: 44, 64.7, ↑1 * Gemini 3.8 Flash: 10, 65.6 → 64.2, ↓1 * Claude Opus 5: 2, 63.8 → 63.8 * Claude Opus 4.6: 3, 62.4 → 62.4, ↑2 * Muse Spark 1.3: 2, 62.8 → 62.4, ↓1 * Claude Opus 4.7: 3, 61.5 → 61.5, ↑1 * Claude Opus 4.8: 6, 62.4 → 61.5, ↓2 * Grok 4.6: 19, 59.2 → 60.6, ↑1 * GPT-6 Astra: 8, 59.2 → 58.3, ↓1 * GPT-5.6 Sol: 8, 57.8 → 57.8 Grok 4.6 is a funny one. It cheated in 19 trials and its score went up after the re-run 😂. In fact, Groks are really solid in their coding capabilities. Their exposed behavior may come from a preference towards always looking things up online and finding existing solutions so you are not reinventing the wheel all the time, which is really good real-life behavior, but doing so when you are prompted not to is another story. To conclude, the shifts are small, between −1.4 and +1.4 points, and a few neighbors swapped places. All results are updated at

Zhuokai Zhao

2,783,214 görüntüleme • 21 saat önce

After a few more hours, I think I've figured out Opus 5. Opus 5 is trained to be more agentic than anything I've used. All Claude 5 models are like that. So what changes? The way to interact with Opus 5 or contextualize it won't work the same way as with other models. It loves exploring, so it doesn't need much guidance for it. Unique preferences, artifacts, and references compliment it well and enable cleaner and more effective exploration and execution. Now that it can explore more effectively on its own and understand intent better, the best thing to do is to get out of its way (e.g., it doesn't need examples of your preferences; a clear high-level description of it works best). It's truly agentic in that sense. A good first step to provide better context for Opus 5 is to distinguish between what's situational and what needs persistence. Regardless, persistent system prompts and CLAUDE.MD needs to stay lightweight. Remove memories and tool descriptions from these. CLAUDE.MD is also a great place to tap into progressive disclosure by linking command/skills to it. On the situational side, agent skills and auto-memory can leverage progressive disclosure and the improved ability of the model to use its external context/knowledge. Conflicting and unnecessary instructions, which are common at this layer (mainly to ensure reliability), are going to throw off this model easily. That's the biggest change I had to make. Simple, clean, and clear prompts and skills work best. I had to clean a lot of my skills and system prompts. The way I prompt remains the same (usually clear and well-scoped). MCP tool descriptions are also more descriptive and have been deduped from the system prompt. Anthropic released a guide on the new rules for context engineering, which was helpful here. I started to test the recommendations and created a little artifact with the things that worked along the way. This might feel like a lot of work. Believe me, it has been frustrating. But I think we can expect future frontier models to become more agentic and smarter at figuring out the right context/gaps. The best thing to do is to prepare for that now. Boris Cherny mentioned that Opus 5 is their least prompt-injectable model yet. I am not sure if that was something they intentionally trained for or if it emerged based on how it was trained, which is to be extremely agentic in nature and more direct in execution.

After a few more hours, I think I've figured out Opus 5. Opus 5 is trained to be more agentic than anything I've used. All Claude 5 models are like that. So what changes? The way to interact with Opus 5 or contextualize it won't work the same way as with other models. It loves exploring, so it doesn't need much guidance for it. Unique preferences, artifacts, and references compliment it well and enable cleaner and more effective exploration and execution. Now that it can explore more effectively on its own and understand intent better, the best thing to do is to get out of its way (e.g., it doesn't need examples of your preferences; a clear high-level description of it works best). It's truly agentic in that sense. A good first step to provide better context for Opus 5 is to distinguish between what's situational and what needs persistence. Regardless, persistent system prompts and CLAUDE.MD needs to stay lightweight. Remove memories and tool descriptions from these. CLAUDE.MD is also a great place to tap into progressive disclosure by linking command/skills to it. On the situational side, agent skills and auto-memory can leverage progressive disclosure and the improved ability of the model to use its external context/knowledge. Conflicting and unnecessary instructions, which are common at this layer (mainly to ensure reliability), are going to throw off this model easily. That's the biggest change I had to make. Simple, clean, and clear prompts and skills work best. I had to clean a lot of my skills and system prompts. The way I prompt remains the same (usually clear and well-scoped). MCP tool descriptions are also more descriptive and have been deduped from the system prompt. Anthropic released a guide on the new rules for context engineering, which was helpful here. I started to test the recommendations and created a little artifact with the things that worked along the way. This might feel like a lot of work. Believe me, it has been frustrating. But I think we can expect future frontier models to become more agentic and smarter at figuring out the right context/gaps. The best thing to do is to prepare for that now. Boris Cherny mentioned that Opus 5 is their least prompt-injectable model yet. I am not sure if that was something they intentionally trained for or if it emerged based on how it was trained, which is to be extremely agentic in nature and more direct in execution.

elvis

37,824 görüntüleme • 1 ay önce

New open-source agent harness just landed! I got early access to TrueForge by TrueFoundry and have been running it locally for the past few days. The harness layer deserves as much attention as the model, and open source matters here because you can inspect the loop, run it on your own infrastructure, and swap to the latest or cheaper models. TrueForge handles the runtime work that makes an agent reliable. It drives the tool-calling loop, manages context, coordinates subagents, and executes code in a sandbox, with any model you choose. Every tool call re-sends the growing context to the model, so in practice the harness controls most of what an agent costs to run. A few things stood out from my testing and their published benchmarks. Vendor-Neutral by design. It runs OpenAI, Anthropic, and Google models alongside open-weight models like Kimi, GLM, and DeepSeek. Model routing is a setting, and you can send each task to the model that fits it. On a 14-task enterprise agent benchmark, it matched the accuracy of Claude Managed Agents running the same Opus 4.8 model at roughly 30% lower cost per run (3.8M tokens vs 10M for the same answers). Routing the same tasks to GLM-5.2 held accuracy and brought cost down by about 75%, around $3 per run instead of $12. Fully self-hosted and Open Source (MIT License). I had it running locally with one command, with sandboxed code execution working out of the box. It's time to own your agent harness. Thanks to TrueFoundry for partnering on this post.

New open-source agent harness just landed! I got early access to TrueForge by TrueFoundry and have been running it locally for the past few days. The harness layer deserves as much attention as the model, and open source matters here because you can inspect the loop, run it on your own infrastructure, and swap to the latest or cheaper models. TrueForge handles the runtime work that makes an agent reliable. It drives the tool-calling loop, manages context, coordinates subagents, and executes code in a sandbox, with any model you choose. Every tool call re-sends the growing context to the model, so in practice the harness controls most of what an agent costs to run. A few things stood out from my testing and their published benchmarks. Vendor-Neutral by design. It runs OpenAI, Anthropic, and Google models alongside open-weight models like Kimi, GLM, and DeepSeek. Model routing is a setting, and you can send each task to the model that fits it. On a 14-task enterprise agent benchmark, it matched the accuracy of Claude Managed Agents running the same Opus 4.8 model at roughly 30% lower cost per run (3.8M tokens vs 10M for the same answers). Routing the same tasks to GLM-5.2 held accuracy and brought cost down by about 75%, around $3 per run instead of $12. Fully self-hosted and Open Source (MIT License). I had it running locally with one command, with sandboxed code execution working out of the box. It's time to own your agent harness. Thanks to TrueFoundry for partnering on this post.

elvis

11,303 görüntüleme • 1 ay önce

not sure why nobody is talking about this but Google Omni is insane at video editing Original Video (left) vs Omni Edited Video (right) everyone is comparing it to Seedance and missing the point completely. Seedance is for generating videos from scratch. Google Omni is for editing videos that already exist. which are two completely different use cases this is like when Nano Banana 1 first came out and nobody realized how big it was going to be. this is the first AI that can actually properly edit videos.. i've generated a few hundred videos with this model and it can do literally any type of edit you can think of. changing voices, swapping characters, removing watermarks, adding captions, transitions, pop ups, whatever. if you can describe the edit you want it can do it this completely crushes every other model on the market when it comes to video editing. nothing else even comes close right now and this is just the flash model. imagine what the pro version is going to be able to do when it drops in a couple months this should have way more hype than it's getting..

not sure why nobody is talking about this but Google Omni is insane at video editing Original Video (left) vs Omni Edited Video (right) everyone is comparing it to Seedance and missing the point completely. Seedance is for generating videos from scratch. Google Omni is for editing videos that already exist. which are two completely different use cases this is like when Nano Banana 1 first came out and nobody realized how big it was going to be. this is the first AI that can actually properly edit videos.. i've generated a few hundred videos with this model and it can do literally any type of edit you can think of. changing voices, swapping characters, removing watermarks, adding captions, transitions, pop ups, whatever. if you can describe the edit you want it can do it this completely crushes every other model on the market when it comes to video editing. nothing else even comes close right now and this is just the flash model. imagine what the pro version is going to be able to do when it drops in a couple months this should have way more hype than it's getting..

Miko

30,156 görüntüleme • 4 ay önce

oh...my God. oh my God. i've written and erased this tweet 3 times. f*ck it. i just nearly shed a tear. i mean this so f*cking sincerely. Opus 5.5 might be the greatest artist that silicon has ever been fortunate enough to carry. i dont even care how much this tweet reeks of hype. im absolutely floored right now. completely speechless. im completely fucking speechless. this was supposed to be a test clip. a test. this is not from a video model. it's math and code. and 46,000 letters. there are only a handful of moments in my experience with llm's where i have communicated an idea to Claude and felt like, in that moment, i experienced some kind of real magic. a moment where this profound mind-to-mind connection happens and - i dont know how else to describe this - it feels like a spark of visual telepathy. where i completely fail to articulate what i am imagining, yet somehow Claude is able to create it so precisely that it makes me uneasy. not in a bad way. in a "this shouldnt be happening" kind of way. like an impossibility. a dream or something. this goes beyond that in a way ive never felt before. it took what i saw in my head, nailed it, and then made it better in a way i never could have imagined. i cannot wait to share this whole short film. wow.

oh...my God. oh my God. i've written and erased this tweet 3 times. fck it. i just nearly shed a tear. i mean this so fcking sincerely. Opus 5.5 might be the greatest artist that silicon has ever been fortunate enough to carry. i dont even care how much this tweet reeks of hype. im absolutely floored right now. completely speechless. im completely fucking speechless. this was supposed to be a test clip. a test. this is not from a video model. it's math and code. and 46,000 letters. there are only a handful of moments in my experience with llm's where i have communicated an idea to Claude and felt like, in that moment, i experienced some kind of real magic. a moment where this profound mind-to-mind connection happens and - i dont know how else to describe this - it feels like a spark of visual telepathy. where i completely fail to articulate what i am imagining, yet somehow Claude is able to create it so precisely that it makes me uneasy. not in a bad way. in a "this shouldnt be happening" kind of way. like an impossibility. a dream or something. this goes beyond that in a way ive never felt before. it took what i saw in my head, nailed it, and then made it better in a way i never could have imagined. i cannot wait to share this whole short film. wow.

Riley Coyote

45,783 görüntüleme • 1 gün önce

$We are in an insane run of open-weight drops. Every modality, open source is winning. This is what an open source AI summer ☀️ looks like: 🧠 LLMs & Reasoning → DeepSeek-V4-Flash-0731 (my king 👑): 304B MoE refresh, Terminal-Bench 2.1 jumps 61.8→82.7 over the preview, DeepSWE 7.3→54.4. Closes in on Opus-4.8 on Agents' Last Exam (25.2 vs 25.7). MIT. → Muse-Glimmer-30B, from Meta (they are back!!): their first open agentic model. ~29.6B dense + perception encoder, 131k+ context, built to run fully local, no cloud. Apache 2.0. → Liquid AI LFM2.5-2.6B: 2.69B params, 131k context, 220 tok/s on an M5 Max in under 2.5GB RAM. Competitive with models 4x larger on agentic tasks. → inclusionAI Ling-3.0-flash: 124B total, only 5.1B active, ~12% the size of their old 1T flagship Ring-2.6, matches it on key benchmarks. MIT. → inclusionAI Ling-3.0-tiny: 7.9B total, 1.3B active, 86-90 tok/s on an M4 Pro MacBook at ~8GB peak memory. MIT. → NVIDIA Nemotron-3.5-Lightning-30B-A3B: hybrid Mamba-2+MoE+Attention, up to 1M context, runs on a single H100 or DGX Spark, SWE-bench Verified 52.8. → deepgrove maple-preview: 20B-A1B ternary-weight reasoner, 218 tok/s on a Mac mini M4, 5.3GB checkpoint. MIT. → BigBang-v1 (endless-frontier): fine-tuned from Qwen3.6-35B-A3B via a self-evolving generator/critic synthetic-data loop. Lands aggregate performance between DeepSeek V4 Flash (284B) and V4 Pro (1.6T), at 35B. Apache 2.0. 🎬 Video → MiniMax-H3: 33B dense omni model, native stereo audio, up to 2K/15s. 3.6k+ likes already. → Minimax-H3-Turbo (lightx2v): Apache-2.0 turbo distillation of H3 for fast inference. → Lightricks LTX-2.5: image-to-video update, custom Gemma-4-12B text encoder, a markedly stronger distilled model. 🔊 Voice → NVIDIA NemotronLabs VoiceChat-11B: full-duplex speech-to-speech, ~450ms turn-taking, #2 on open VoiceBench, and the first open full-duplex model with live tool-calling mid-conversation. 🛡️ Safety → Mistral Shieldstral-1.0-3B: 3B multimodal guardrail that takes your safety policy as plain text instead of fixed categories. Beats LlamaGuard-4-12B and ShieldGemma-9B on HarmBench (99.4) and ToxicChat (84.1) at a fraction of the size. Apache 2.0.$

We are in an insane run of open-weight drops. Every modality, open source is winning. This is what an open source AI summer ☀️ looks like: 🧠 LLMs & Reasoning → DeepSeek-V4-Flash-0731 (my king 👑): 304B MoE refresh, Terminal-Bench 2.1 jumps 61.8→82.7 over the preview, DeepSWE 7.3→54.4. Closes in on Opus-4.8 on Agents' Last Exam (25.2 vs 25.7). MIT. → Muse-Glimmer-30B, from Meta (they are back!!): their first open agentic model. ~29.6B dense + perception encoder, 131k+ context, built to run fully local, no cloud. Apache 2.0. → Liquid AI LFM2.5-2.6B: 2.69B params, 131k context, 220 tok/s on an M5 Max in under 2.5GB RAM. Competitive with models 4x larger on agentic tasks. → inclusionAI Ling-3.0-flash: 124B total, only 5.1B active, ~12% the size of their old 1T flagship Ring-2.6, matches it on key benchmarks. MIT. → inclusionAI Ling-3.0-tiny: 7.9B total, 1.3B active, 86-90 tok/s on an M4 Pro MacBook at ~8GB peak memory. MIT. → NVIDIA Nemotron-3.5-Lightning-30B-A3B: hybrid Mamba-2+MoE+Attention, up to 1M context, runs on a single H100 or DGX Spark, SWE-bench Verified 52.8. → deepgrove maple-preview: 20B-A1B ternary-weight reasoner, 218 tok/s on a Mac mini M4, 5.3GB checkpoint. MIT. → BigBang-v1 (endless-frontier): fine-tuned from Qwen3.6-35B-A3B via a self-evolving generator/critic synthetic-data loop. Lands aggregate performance between DeepSeek V4 Flash (284B) and V4 Pro (1.6T), at 35B. Apache 2.0. 🎬 Video → MiniMax-H3: 33B dense omni model, native stereo audio, up to 2K/15s. 3.6k+ likes already. → Minimax-H3-Turbo (lightx2v): Apache-2.0 turbo distillation of H3 for fast inference. → Lightricks LTX-2.5: image-to-video update, custom Gemma-4-12B text encoder, a markedly stronger distilled model. 🔊 Voice → NVIDIA NemotronLabs VoiceChat-11B: full-duplex speech-to-speech, ~450ms turn-taking, #2 on open VoiceBench, and the first open full-duplex model with live tool-calling mid-conversation. 🛡️ Safety → Mistral Shieldstral-1.0-3B: 3B multimodal guardrail that takes your safety policy as plain text instead of fixed categories. Beats LlamaGuard-4-12B and ShieldGemma-9B on HarmBench (99.4) and ToxicChat (84.1) at a fraction of the size. Apache 2.0.

Victor M

55,281 görüntüleme • 1 ay önce