正在加载视频...

视频加载失败

加载此视频时出现问题。这可能是由于临时网络问题，或视频可能不可用。

Black hole visualisation by MiniMax M2.5 Holy Shit i have never seen this output 😮 >SOTA performance better than Opus4.6 at coding SWE-Bench Verified 80.2%, BrowseComp 76.3% >$1 per hour with 100 tps i.e 3 times the speed of Opus4.6 > Only 10B active parameter

Chetaslua

17,328 subscribers

17,725 次观看 • 6 个月前 •via X (Twitter)

Anya Rossi• Live Now

Private livecam show

0 条评论

暂无评论

原始帖子的评论将显示在这里

相关视频

MiniMax M2.5 > Claude Opus 4.6 I have been using it for a couple of days and have never hit the limit. I am currently on the Coding Plus plan which gives me 300 prompts per 5 hours. My new setup is OpenCode + MiniMax M2.5 What I have noticed: M2.5 – handles multi-step coding workflows and repo-level reasoning more reliably – better at chaining actions. think: generate test, fix bug, refactor, repeat – high TPS throughput M2.5 hits ~80%+ on SWE-Bench coding benchmarks on par with Opus and runs these tasks 37% faster than its predecessor while costing ~1/10th - 1/20th the price. Check out the video.

MiniMax M2.5 > Claude Opus 4.6 I have been using it for a couple of days and have never hit the limit. I am currently on the Coding Plus plan which gives me 300 prompts per 5 hours. My new setup is OpenCode + MiniMax M2.5 What I have noticed: M2.5 – handles multi-step coding workflows and repo-level reasoning more reliably – better at chaining actions. think: generate test, fix bug, refactor, repeat – high TPS throughput M2.5 hits ~80%+ on SWE-Bench coding benchmarks on par with Opus and runs these tasks 37% faster than its predecessor while costing ~1/10th - 1/20th the price. Check out the video.

Pratham

184,716 次观看 • 5 个月前

Holy Shit This is insane result 🤯 I used opus 4.6 as a reviewer/planner and 4 worker minimax M2.7 agent and this the result Voxel art of eiffel tower < loop set to 5 that means opus will tell MiniMax (official) to make it better 5 times >

Holy Shit This is insane result 🤯 I used opus 4.6 as a reviewer/planner and 4 worker minimax M2.7 agent and this the result Voxel art of eiffel tower < loop set to 5 that means opus will tell MiniMax (official) to make it better 5 times >

Chetaslua

82,988 次观看 • 4 个月前

Kimi K2.6 is godly in terms of webdev and still SOTA chinese model 😼 > best in deep SWE bench even after launches of multiple models from different chinese labs Look at the watery feel one shotted it < using custom harness - 4 kimi in parallel just like grok >

Kimi K2.6 is godly in terms of webdev and still SOTA chinese model 😼 > best in deep SWE bench even after launches of multiple models from different chinese labs Look at the watery feel one shotted it < using custom harness - 4 kimi in parallel just like grok >

Chetaslua

117,033 次观看 • 2 个月前

Seedance 2.5 is mythos moment of Video Models > more than 30 seconds > 4k resolution > prompt adherence is on another level that we never seen before > our reports was right , this is why follow us and join our server > July will be do much fun

Seedance 2.5 is mythos moment of Video Models > more than 30 seconds > 4k resolution > prompt adherence is on another level that we never seen before > our reports was right , this is why follow us and join our server > July will be do much fun

Chetaslua

207,536 次观看 • 1 个月前

🚨 Deepseek V4 pro < api is sus> i have read multiple reports on x that it is routing to fable 5 , if it is right then we can see the jump in performance normal If it is not routing to fable 5 < quite impossible > then they made better model than kimi k3 , here is output

🚨 Deepseek V4 pro < api is sus> i have read multiple reports on x that it is routing to fable 5 , if it is right then we can see the jump in performance normal If it is not routing to fable 5 < quite impossible > then they made better model than kimi k3 , here is output

Chetaslua

99,585 次观看 • 26 天前

🚨Gemini 3 Pro Greatness Thread > This thread will showcase creativity, world knowledge, coding capabilities and scientific knowledge 1. Solar System Simulation by Chetas

🚨Gemini 3 Pro Greatness Thread > This thread will showcase creativity, world knowledge, coding capabilities and scientific knowledge 1. Solar System Simulation by Chetas

Chetaslua

40,254 次观看 • 10 个月前

🚨 Kimi K2.7-Code is actually insane >#1 on Program Bench >#1 on MLS-Bench Lite (multi-step ML research) >Refreshing to see honest benchmarks instead of cherry-picked After claude fable debacle i want OSS to win FR to save us from dystopian end

🚨 Kimi K2.7-Code is actually insane >#1 on Program Bench >#1 on MLS-Bench Lite (multi-step ML research) >Refreshing to see honest benchmarks instead of cherry-picked After claude fable debacle i want OSS to win FR to save us from dystopian end

Chetaslua

37,707 次观看 • 2 个月前

🚨 Gemini 3.0 Flash > one shotted this trip on globe planner > super fast and to the point > i think this will be beast in gemini cli for faster and to the point coding

🚨 Gemini 3.0 Flash > one shotted this trip on globe planner > super fast and to the point > i think this will be beast in gemini cli for faster and to the point coding

Chetaslua

36,421 次观看 • 10 个月前

🚨Gemini 3.0 Pro - ECP. one shot in 2 Min Holy Shit Guys Gemini made GAMEBOY with original games This is literally my favourite the level of detail is beyond any LLM till date , this is the best output i have ever seen Prompt and code pen link in thread

🚨Gemini 3.0 Pro - ECP. one shot in 2 Min Holy Shit Guys Gemini made GAMEBOY with original games This is literally my favourite the level of detail is beyond any LLM till date , this is the best output i have ever seen Prompt and code pen link in thread

Chetaslua

63,440 次观看 • 10 个月前

Kimi K3 absolutely mogs every open source model this is on par , head on head all case i tried in coding with fable , i feel weird to write , but sometimes it output better quality than fable too I don't know how to comprehend this , i have lots of test and comparison 😼

Kimi K3 absolutely mogs every open source model this is on par , head on head all case i tried in coding with fable , i feel weird to write , but sometimes it output better quality than fable too I don't know how to comprehend this , i have lots of test and comparison 😼

Chetaslua

130,851 次观看 • 1 个月前

Grok should be praised a lot , this is one of the biggest launch > priced way less than even kimi k2.7 , glm 5.2 > on artificial code analysis only behind fable 5 and gpt 5.5 xhigh > this is well rounded launch i will showcase a 3d test that will blow everyone mind

Grok should be praised a lot , this is one of the biggest launch > priced way less than even kimi k2.7 , glm 5.2 > on artificial code analysis only behind fable 5 and gpt 5.5 xhigh > this is well rounded launch i will showcase a 3d test that will blow everyone mind

Chetaslua

25,644 次观看 • 1 个月前

🚨 Deepseek V4 pro GA is being tested to few users > this is the output on the same prompt that was used for kimi k3 and GPT sol 5.6 in the quoted video > The details are solid , addition of tents , campfire and everything this model looks promising

🚨 Deepseek V4 pro GA is being tested to few users > this is the output on the same prompt that was used for kimi k3 and GPT sol 5.6 in the quoted video > The details are solid , addition of tents , campfire and everything this model looks promising

Chetaslua

172,573 次观看 • 27 天前

🚨 GPT 5.6 Pro first output on the same prompt we are getting started > frontend/ webdev is not solved or improved yet > but understanding increased a lot > it started to take 20-40 mins again like it used to do before 5.5 pro

🚨 GPT 5.6 Pro first output on the same prompt we are getting started > frontend/ webdev is not solved or improved yet > but understanding increased a lot > it started to take 20-40 mins again like it used to do before 5.5 pro

Chetaslua

450,779 次观看 • 1 个月前

🚨 Gemini 3.0 Pro > google is winning very hard and mark my word this will be the best front end and web dev model ever > i will test more prompts on this upcoming model > i was inactive sorry for my absence guys

🚨 Gemini 3.0 Pro > google is winning very hard and mark my word this will be the best front end and web dev model ever > i will test more prompts on this upcoming model > i was inactive sorry for my absence guys

Chetaslua

511,632 次观看 • 10 个月前

Opus 5 is 3d beast On my Opus i swear i have never seen this level of detail in one shot ever in my life even multiple shots cant produce this , i am shaking with this out holy opus , btw its token hungry in 3 prompt ate entire usage even fable 5 is not that token hungry

Opus 5 is 3d beast On my Opus i swear i have never seen this level of detail in one shot ever in my life even multiple shots cant produce this , i am shaking with this out holy opus , btw its token hungry in 3 prompt ate entire usage even fable 5 is not that token hungry

Chetaslua

620,518 次观看 • 22 天前

🚨 Gemini 3.5 Pro < old checkpoint > Gojo Satoru Aura Farming > Cursed technique simulation coded by it one shot Gemini as always surprise with insane results when it comes to taste and world knowledge < just solve laziness and bring back personality of 2.5 pro >

🚨 Gemini 3.5 Pro < old checkpoint > Gojo Satoru Aura Farming > Cursed technique simulation coded by it one shot Gemini as always surprise with insane results when it comes to taste and world knowledge < just solve laziness and bring back personality of 2.5 pro >

Chetaslua

14,925 次观看 • 2 个月前

🚨 Gemini 3.5 Flash speed numbers just dropped 3-9x faster than current Flash peak speeds over 900 tokens/sec capability hasn't dropped, actually improved parameter count estimated larger than expected video: full shooter game one-shot. holds up at speed. Also currently testing 3.5 Pro on Arena. Should I start posting Pro outputs now or wait for public launch? vote below 🫡

🚨 Gemini 3.5 Flash speed numbers just dropped 3-9x faster than current Flash peak speeds over 900 tokens/sec capability hasn't dropped, actually improved parameter count estimated larger than expected video: full shooter game one-shot. holds up at speed. Also currently testing 3.5 Pro on Arena. Should I start posting Pro outputs now or wait for public launch? vote below 🫡

Chetaslua

41,991 次观看 • 2 个月前

Alibaba just released a coding model that hits 82 percent on SWE-Bench Verified. That is the highest score ever published for an open-source model. The weights are free. The license is Apache 2.0. You can run it today. The model is Qwen 4 Coder 32B. Here is what 82 percent on SWE-Bench Verified actually means. SWE-Bench Verified tests whether an AI can autonomously resolve real bugs pulled from real production GitHub repositories. Not synthetic exercises. Real open-source projects that real teams depend on. A model gets a bug report, reads the code, writes a fix, and either passes the test suite or it does not. At 82 percent, Qwen 4 Coder 32B resolves 82 out of every 100 real production bugs it is given. Without a human guiding it. On code it has never seen before. For comparison: Qwen 4 Coder 32B: 82 percent SWE-Bench Verified. Open source. Apache 2.0. Claude Fable 5: 80.3 percent SWE-Bench Pro. $10 input / $50 output per million tokens. Currently suspended. GPT-5.6 Sol: Competitive on Terminal-Bench. $5 input / $30 output per million tokens. An open-weight model that you can download and run for free just beat both of them on the benchmark designed to measure real software engineering capability. Here is the architecture. Qwen 4 Coder 32B is a 32 billion parameter dense model. Not a Mixture-of-Experts. Every parameter is active on every request. This matters for inference: a dense 32B model runs on 22 gigabytes of VRAM, which fits on a single high-end consumer GPU or a MacBook Pro with 64GB of unified memory. The smaller variant, Qwen 4 Coder 4B, runs at approximately 135 tokens per second on an M5 Max and fits inside 8 gigabytes of RAM. For a model with usable coding capability, that is a new bar for what fits in a single laptop. The training methodology continued Alibaba's approach of reinforcement learning on verifiable coding tasks. The model gets rewarded when its code passes tests. It gets penalized when it fails. Over millions of training steps, the model learns to write code that actually runs rather than code that looks plausible. License: Apache 2.0. Full commercial use. No attribution requirement. No revenue threshold. No monthly active user ceiling. Weights: Hugging Face, available today. Runs on: vLLM, Ollama, SGLang, and any standard GGUF-compatible inference engine. Qwen 4 32B also runs at approximately 135 tokens per second on an M5 Max chip, setting a new bar for what a sub-8GB model can do on Apple Silicon. The open-source coding model just beat the best closed-source model in the world on the benchmark designed to test whether AI can actually do software engineering. The weights are free. The subscription is optional. Source: Autom8Labs AI Insight July 2026, State of Open Source LLMs June 2026, Kunal Ganglani blog June 2026.

Alibaba just released a coding model that hits 82 percent on SWE-Bench Verified. That is the highest score ever published for an open-source model. The weights are free. The license is Apache 2.0. You can run it today. The model is Qwen 4 Coder 32B. Here is what 82 percent on SWE-Bench Verified actually means. SWE-Bench Verified tests whether an AI can autonomously resolve real bugs pulled from real production GitHub repositories. Not synthetic exercises. Real open-source projects that real teams depend on. A model gets a bug report, reads the code, writes a fix, and either passes the test suite or it does not. At 82 percent, Qwen 4 Coder 32B resolves 82 out of every 100 real production bugs it is given. Without a human guiding it. On code it has never seen before. For comparison: Qwen 4 Coder 32B: 82 percent SWE-Bench Verified. Open source. Apache 2.0. Claude Fable 5: 80.3 percent SWE-Bench Pro. $10 input / $50 output per million tokens. Currently suspended. GPT-5.6 Sol: Competitive on Terminal-Bench. $5 input / $30 output per million tokens. An open-weight model that you can download and run for free just beat both of them on the benchmark designed to measure real software engineering capability. Here is the architecture. Qwen 4 Coder 32B is a 32 billion parameter dense model. Not a Mixture-of-Experts. Every parameter is active on every request. This matters for inference: a dense 32B model runs on 22 gigabytes of VRAM, which fits on a single high-end consumer GPU or a MacBook Pro with 64GB of unified memory. The smaller variant, Qwen 4 Coder 4B, runs at approximately 135 tokens per second on an M5 Max and fits inside 8 gigabytes of RAM. For a model with usable coding capability, that is a new bar for what fits in a single laptop. The training methodology continued Alibaba's approach of reinforcement learning on verifiable coding tasks. The model gets rewarded when its code passes tests. It gets penalized when it fails. Over millions of training steps, the model learns to write code that actually runs rather than code that looks plausible. License: Apache 2.0. Full commercial use. No attribution requirement. No revenue threshold. No monthly active user ceiling. Weights: Hugging Face, available today. Runs on: vLLM, Ollama, SGLang, and any standard GGUF-compatible inference engine. Qwen 4 32B also runs at approximately 135 tokens per second on an M5 Max chip, setting a new bar for what a sub-8GB model can do on Apple Silicon. The open-source coding model just beat the best closed-source model in the world on the benchmark designed to test whether AI can actually do software engineering. The weights are free. The subscription is optional. Source: Autom8Labs AI Insight July 2026, State of Open Source LLMs June 2026, Kunal Ganglani blog June 2026.

Harman

41,278 次观看 • 1 个月前