Video wird geladen...

Video konnte nicht geladen werden

Zur Startseite

Qwen 3.8 Max on my shader test. Basic impression after a bunch of experiments is that it is a solid model, but not Kimi K3 level in my experience so far.

emollick's profile picture

Ethan Mollick

378,946 subscribers

39,760 Aufrufe • vor 1 Monat •via X (Twitter)

Wissenschaft & TechnologieBildung
LIVE

Anya Rossi• Live Now

Private livecam show

0 Kommentare

Keine Kommentare verfügbar

Kommentare vom Original-Post werden hier angezeigt

Ähnliche Videos

I tested Qwen 3.8 Max vs Kimi K3 Same scene, lighting, props, detail requirements. Qwen's output is decent, but it's not close to Kimi K3. Qwen's output came out messy, text handling was off, and the composition felt unorganized next to Kimi's version. In my opinion, Qwen isn't second to Fable 5 either, and it's also below GPT-5.6 Sol. It's a solid model, but not at that tier. Kimi K3 is still clearly ahead here.
0:13

I tested Qwen 3.8 Max vs Kimi K3 Same scene, lighting, props, detail requirements. Qwen's output is decent, but it's not close to Kimi K3. Qwen's output came out messy, text handling was off, and the composition felt unorganized next to Kimi's version. In my opinion, Qwen isn't second to Fable 5 either, and it's also below GPT-5.6 Sol. It's a solid model, but not at that tier. Kimi K3 is still clearly ahead here.

Bhavy☄️

34,045 Aufrufe • vor 2 Monaten

Kimi K3 on my shader test: "create a visually interesting shader that can run in twigl-dot-app make it like an infinite city of neo-gothic towers partially drowned in a stormy ocean with large waves." "Make it better" Very good model, not Sol Max or Fable, but great open weights
0:13

Kimi K3 on my shader test: "create a visually interesting shader that can run in twigl-dot-app make it like an infinite city of neo-gothic towers partially drowned in a stormy ocean with large waves." "Make it better" Very good model, not Sol Max or Fable, but great open weights

Ethan Mollick

280,576 Aufrufe • vor 2 Monaten

To show you how beast of a model is opus 5 this sol 5.6 against it that wall where opus is shoooting great wall of china < kimi k3 , qwen 3.8 > we are so back , wait for mac os code to compare with kimi k3
0:11

To show you how beast of a model is opus 5 this sol 5.6 against it that wall where opus is shoooting great wall of china < kimi k3 , qwen 3.8 > we are so back , wait for mac os code to compare with kimi k3

Chetaslua

171,146 Aufrufe • vor 2 Monaten

Qwen 3.8 Max is actually impressive Sent a sky pic to it along with Claude Opus 5, Kimi K3, and GPT 5.6 Sol, and asked them to draw an animal silhouette based on the cloud shape Qwen 3.8 Max is honestly on par with all the other models
0:11

Qwen 3.8 Max is actually impressive Sent a sky pic to it along with Claude Opus 5, Kimi K3, and GPT 5.6 Sol, and asked them to draw an animal silhouette based on the cloud shape Qwen 3.8 Max is honestly on par with all the other models

Ann Nguyen

473,353 Aufrufe • vor 1 Monat

🚨 Qwen new checkpoint - kieran As usual it claims he is claude and made by anthropic but with tokenizer prompt , its clear - QWEN model Ignore website this is my test to know which model is which behind the scene , should i test it and compare with kimi k3 / opus 5 /sol 5.6
0:16

🚨 Qwen new checkpoint - kieran As usual it claims he is claude and made by anthropic but with tokenizer prompt , its clear - QWEN model Ignore website this is my test to know which model is which behind the scene , should i test it and compare with kimi k3 / opus 5 /sol 5.6

Chetaslua

15,544 Aufrufe • vor 2 Monaten

Okay, I am fairly confident in my hypothesis now. The secret sauce behind Qwen 3.8 27B becomes almost immediately evident during testing. It is not the training data. In fact, I doubt any SFT was involved at all. The model was simply allowed GRPO with a more liberal reasoning context length. As I have explained to you before, distillation through GRPO is powerful, and in certain contexts vastly superior to K/L divergence distillation. Qwen used the outputs of Qwen 3.8 Max as the reinforcement learning objective for Qwen 3.8 27B and told the model reason as much as it has to in order to match the output. Remember, Intelligence is solution, not compression. This is like nesting the output of a function as an additional input to itself, and it is absolutely brilliant.
0:17

Okay, I am fairly confident in my hypothesis now. The secret sauce behind Qwen 3.8 27B becomes almost immediately evident during testing. It is not the training data. In fact, I doubt any SFT was involved at all. The model was simply allowed GRPO with a more liberal reasoning context length. As I have explained to you before, distillation through GRPO is powerful, and in certain contexts vastly superior to K/L divergence distillation. Qwen used the outputs of Qwen 3.8 Max as the reinforcement learning objective for Qwen 3.8 27B and told the model reason as much as it has to in order to match the output. Remember, Intelligence is solution, not compression. This is like nesting the output of a function as an additional input to itself, and it is absolutely brilliant.

Aesther

122,746 Aufrufe • vor 1 Monat

DeepSeek V4 Flash being ranked almost equal to Opus 4.8 is actually insane to me. Against Kimi K3, Opus 5, GPT-5.6 Sol, and Qwen 3.8 Max Preview, you give up a LOT by choosing DeepSeek. This is not a 2-point difference in practice. I’ll post the Opus 4.8 comparison next because you need to see this.
0:19

DeepSeek V4 Flash being ranked almost equal to Opus 4.8 is actually insane to me. Against Kimi K3, Opus 5, GPT-5.6 Sol, and Qwen 3.8 Max Preview, you give up a LOT by choosing DeepSeek. This is not a 2-point difference in practice. I’ll post the Opus 4.8 comparison next because you need to see this.

OmedTheVibeCoder

26,990 Aufrufe • vor 1 Monat

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.
0:13

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 Aufrufe • vor 2 Monaten

Codex can run Qwen-3.8-max now as well!! Alibaba most capable model, dropped today and it's already in my codex picker. It's a token plan subscription, not metered api billing. You take the key from your Qwen plan, drop it into Codex Router, and it spends down the plan instead of your card. There's a catch though. Qwen's official setup switches your whole codex over to them, so your ChatGPT models stop showing up at all. That's exactly what Codex Router is for. It adds models to the list instead of replacing them, so sol, Grok, kimi, Deepseek and now Qwen 3.8 max all sit in the same picker and it can grab whichever one suits the job. Router's open source, setup's in the video 👇
0:16

Codex can run Qwen-3.8-max now as well!! Alibaba most capable model, dropped today and it's already in my codex picker. It's a token plan subscription, not metered api billing. You take the key from your Qwen plan, drop it into Codex Router, and it spends down the plan instead of your card. There's a catch though. Qwen's official setup switches your whole codex over to them, so your ChatGPT models stop showing up at all. That's exactly what Codex Router is for. It adds models to the list instead of replacing them, so sol, Grok, kimi, Deepseek and now Qwen 3.8 max all sit in the same picker and it can grab whichever one suits the job. Router's open source, setup's in the video 👇

Ziwen

419,031 Aufrufe • vor 1 Monat

Kimi K3 is now live in LM Studio Bionic! A massive, open frontier model: 2.8T MoE with 1M context window. Kimi K3 in Bionic is hosted on US-based servers, with ZDR enabled by default.
0:10

Kimi K3 is now live in LM Studio Bionic! A massive, open frontier model: 2.8T MoE with 1M context window. Kimi K3 in Bionic is hosted on US-based servers, with ZDR enabled by default.

LM Studio

133,008 Aufrufe • vor 2 Monaten

Kimi K3 is now available in June in anonymous mode. Frontier level AI, in a private workspace for your agent. Try it:
0:15

Kimi K3 is now available in June in anonymous mode. Frontier level AI, in a private workspace for your agent. Try it:

OpenSoftware

54,891 Aufrufe • vor 2 Monaten

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!
0:18

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 Aufrufe • vor 1 Monat

A quick cost per task comparison of Kimi K3 Kimi.ai and GPT-5.6 Sol by OpenAI. At 3.2x cheaper, Kimi K3's model outputs web apps not just on par for design quality, but better across certain domains. Kimi K3 establishes a new design SOTA and Pareto frontier for cost for design tasks.
0:12

A quick cost per task comparison of Kimi K3 Kimi.ai and GPT-5.6 Sol by OpenAI. At 3.2x cheaper, Kimi K3's model outputs web apps not just on par for design quality, but better across certain domains. Kimi K3 establishes a new design SOTA and Pareto frontier for cost for design tasks.

Design Arena

22,584 Aufrufe • vor 2 Monaten

Kimi K3 is a wildly creative model. I asked it to design Dieter Rams' Braun T3 radio in adam. It wrote the CAD, then decided the STEP wasn't enough, piped it into Blender on my Mac and lit a full product shoot in three colorways.
0:10

Kimi K3 is a wildly creative model. I asked it to design Dieter Rams' Braun T3 radio in adam. It wrote the CAD, then decided the STEP wasn't enough, piped it into Blender on my Mac and lit a full product shoot in three colorways.

Zach Dive

24,322 Aufrufe • vor 2 Monaten

What's strange to me is that there's a bunch of really sick visual experiments being achieved with Three.js /tsl, but very few actual games. This is one of the best ones I've seen so far.. Give me more examples!
0:18

What's strange to me is that there's a bunch of really sick visual experiments being achieved with Three.js /tsl, but very few actual games. This is one of the best ones I've seen so far.. Give me more examples!

Gary Simon

42,068 Aufrufe • vor 4 Monaten

"I'm not a human." Fed it to Qwen 3.5 0.8B running locally on my Mac Studio M2 Ultra. It solved it. The CAPTCHA is fake. But sending images to the local model? Very real. I'm not breaking the internet. Yet.
0:12

"I'm not a human." Fed it to Qwen 3.5 0.8B running locally on my Mac Studio M2 Ultra. It solved it. The CAPTCHA is fake. But sending images to the local model? Very real. I'm not breaking the internet. Yet.

stevibe

69,433 Aufrufe • vor 6 Monaten

Been using Qwen 3.8 27B (Q4) locally on 64GB of VRAM. Here is the verdict: SLOW 18 tps with ZERO system prompt to process and that degrades significantly with a harness system prompt and as the context window grows. RIP if you have to compact. I had it implement this PRD and it's been running for 6 hours. By comparison Grok 4.6 and Kimi K3 hosted finished in about ~30 minutes. High hopes, but these 27B variants are too dense. This is not a consumer grade local model - and I consider consumer grade to be anything up to $5000.
0:17

Been using Qwen 3.8 27B (Q4) locally on 64GB of VRAM. Here is the verdict: SLOW 18 tps with ZERO system prompt to process and that degrades significantly with a harness system prompt and as the context window grows. RIP if you have to compact. I had it implement this PRD and it's been running for 6 hours. By comparison Grok 4.6 and Kimi K3 hosted finished in about ~30 minutes. High hopes, but these 27B variants are too dense. This is not a consumer grade local model - and I consider consumer grade to be anything up to $5000.

Burke Holland

122,414 Aufrufe • vor 1 Monat

Just so we’re clear, the OP does NOT include ANY description/hashtags whatsoever. It wasn’t until AFTER they got a bunch of pushback on how ludicrous this is that they went in and added a screenshot of hashtags, which again, is not included on this video but which they claim is the same video.
0:10

Just so we’re clear, the OP does NOT include ANY description/hashtags whatsoever. It wasn’t until AFTER they got a bunch of pushback on how ludicrous this is that they went in and added a screenshot of hashtags, which again, is not included on this video but which they claim is the same video.

Mel

69,856 Aufrufe • vor 4 Monaten

DeepSeek V4 Flash 0731 first test. DeepSeek V4 Flash 0731 is ridiculously cheap compared to Kimi K3, GPT-5.6, and Opus 5 For its price, this is a surprisingly capable model. First challenge: Generate a railway station in Three.js. The result exceeded my expectations for such a low-cost model. It’s definitely one to watch if you’re building 3D projects on a tight budget.
0:15

DeepSeek V4 Flash 0731 first test. DeepSeek V4 Flash 0731 is ridiculously cheap compared to Kimi K3, GPT-5.6, and Opus 5 For its price, this is a surprisingly capable model. First challenge: Generate a railway station in Three.js. The result exceeded my expectations for such a low-cost model. It’s definitely one to watch if you’re building 3D projects on a tight budget.

FHILY👑

21,825 Aufrufe • vor 1 Monat