Took some collaboration w/ @ChatGPT Codex, but finally worked out a pretty good pipeline for my AI Influencer project for @cutparty.
Created for free on my MBP M4 Max 128GB w/ @MiniMax_AI H3 via @PhospheneAI in Pinokio App. Thanks to @AIBizarrothe@cocktailpeanut FOSS!
More๐
๐ข Meet Qwen4-Max, your new AI teammate for work, creativity, and discovery!
Qwen4-Max is built on a new MoE architecture with native vision and a 1M-token context window.
Highlights:
- Outperforms Qwen3.8-Max with up to ~50% fewer tokens, while average cost per task drops over 60% in our internal evaluation.
- Advances real-world productivity through greater proficiency with professional tools, stronger long-horizon execution, and higher-quality deliverables.
- Handles real tasks, including 3D modeling, game development, manufacturing, finance, law, healthcare, and frontier scientific discovery.
- $1.5/M tokens for input, $4.5/Mย tokens for output, and $0.075/M tokens forย implicitย cachingย โ allย lower than Qwen3.8-Max, withย implicit cachingย downย by 70%.
The API is live on QwenCloud. You can also try Qwen4-Max on Qwen Studio, QwenWork and Qoder.
Start building today! ๐
- Blog: https://t.co/AuiWxugjE6
- API: https://t.co/aS1QEkS8nu
- Token Plan: https://t.co/YRvcGdB9Bv
More to come! The open weights of smaller Qwen4 models will be released. Stay tuned. ๐
Thanks AA! It's the first time a model has ranked ahead of Suno's latest release on this chart.
This is only the beginning. We'll open-source the model weights in the next few weeks. We're making the final tweaks so you can run it locally.
And yes, an even more powerful model is on the way.
Stay tuned, peace โ๏ธ
ACE Music v1 just hit #1 on @ArtificialAnlys' music leaderboard. It's the first model on this chart to outrank Suno's latest release.
Try it free at https://t.co/tPBsxSpUXU.
We'll open-source the model weights in the next few weeks. We're making the final tweaks so you can run it locally.
ACE Music v1 debuts at #1 on AA-Music-Vocal v1.1, ahead of Suno v6
ACE Music v1 is the new music model from ACE Studio. Founded in 2018, ACE Studio began in AI singing voice synthesis: its ACE Studio app is an AI music workstation that turns MIDI and lyrics into vocals for music producers. Together with StepFun, it also co-develops ACE-Step, a family of open weights music generation models.
ACE Music v1 generates full tracks with vocals or instrumentals from a text prompt, starting from a description or your own lyrics.
In our testing, it does best on Reggae / Dancehall and Classical / Orchestral prompts.
ACE Music v1 has no API. It is available in the ACE Music app, with a free tier of 20 credits a day and paid plans from $7.90 a month billed yearly, for up to 500 songs a month.
Congratulations to @acemusicAI on the release!
See below for comparisons between ACE Music v1 and other leading models in the Artificial Analysis Music Arena ๐งต
So I tested Fable 5.2 on real work. 2 repos, 105 bugs. Find and fix what you can.
The results:
- Fable 5.2 (max): 58 for $139.65
- Sonnet 5.5 (max): 51.3 for $154.03
- GPT-6 Astra (max): 45 for $33.03
- GPT-6.1 Sol (max): 44.3 for $5.88
- Fable 5.1 (max): 43 for $87.18 (n=1)
- Opus 5.5 (max): 41.7 for $58.53
All three Fable 5.2 runs beat Sonnet 5.5's average. And it got there with less than half of Sonnet's turns (707 per run vs 1,497), for less money.
More effort levels dropping in this thread soon ๐งต
In one of my favorite experiments, I told Fable 5.2 to build an endless, hands-off journey through 3D worlds linked by portals.
25,000 agents created over 450 worlds inspired by books, art, places, all in code.
Play: https://t.co/pu1gO2sufk
Github: https://t.co/NJKDlib6Xp
Claude Fable 5.2 is now #1 on AA-3D, our benchmark for how well AI models build interactive 3D experiences
Claude Fable 5.2 (Max) scores 1271 Elo, 67 points ahead of Claude Sonnet 5.5 (1204) and Claude Opus 5.5 (1202). Its 95% confidence interval (1246 to 1297) does not overlap any other models, so its lead is outside the margin of error.
Key results:
โค It scores 132 points above its predecessor, Claude Fable 5.1 (1139)
โค It is ahead of the highest-ranked OpenAI and Google models on AA-3D, GPT-6 Astra (1168) and Gemini 4 Argon (1147)
More results and scenes in the thread below
During pre-release testing, Claude Fable 5.2 and Claude Mythos 5.2 worked on ambitious projects in many different fields. Here are a few.
From a one-paragraph prompt, Fable 5.2 designed three working electronic devices, including this home voice assistant hub.
Anthropic has released Claude Fable 5.2, taking the #1 spot on the Artificial Analysis Intelligence Index v5 at 65, led by gains in science, knowledge, and coding
At max effort it scores 65, 5 points ahead of Gemini 4 Carbon (high, 60) and 6 ahead of Claude Opus 5.5 (max, 59). Gains are driven by scientific reasoning, knowledge and FrontierCode 2.0, and roughly matching Opus 5.5 on AA-Briefcase and Terminal-Bench 4.0.
Priced at $8/$40 per 1M input/output tokens, a 20% reduction from Fable 5.1, Claude Fable 5.2 averages a Cost per Task of $26.5 with max effort, a ~2.6x increase compared to Opus 5.5 (max, $10.0) and ~2x compared to Fable 5.1 (max, $12.99)
Key takeaways:
โค Claude Fable 5.2 leads 5 of the 9 evaluations in the Intelligence Index: At max effort, Fable 5.2 scores 65, ahead of Gemini 4 Carbon (high, 60), Claude Opus 5.5 (max, 59) and GPT-6 Astra (max, 57), and ~15 points above Claude Fable 5.1 (max, 49). It sets the highest scores on HLE-Diamond (80%, previous best 69% from Gemini 4 Carbon), AA-Omniscience v1.1 (67, previous best 54 from Gemini 4 Carbon), Terminal-Bench-Science (73%, previous best 65% from Gemini 4 Carbon), CritPt v2 hard subset (77%, previous best 72% from GPT-6 Astra) and FrontierCode 2.0 (49%, previous best 46% from Claude Opus 5.5)
โค Less progress in agentic knowledge work and terminal coding: Claude Fable 5.2 improves on Fable 5.1 across capabilities, but sometimes only to Opus 5.5 levels. It reaches 1785 Elo on AA-Briefcase, within the confidence interval of Opus 5.5 (1800) and just behind Claude Sonnet 5.5 (1813). On GDP.pdf it reaches 24% all-pass against 26% for Opus 5.5. On Terminal-Bench 4.0 it scores 60%, level with Opus 5.5 and behind Kimi K3.1 Pro (67%), Gemini 4 Carbon (65%) and Sonnet 5.5 (64%). AutomationBench-AA is an exception, at 76% against 70% for Opus 5.5, narrowly behind Gemini 4 Carbon, 78%
โค Lower efforts still take leading positions: With high and xhigh effort, Claude Fable 5.2 scores 62, ahead of every other model at any effort, using 98k output tokens per task against 173k for Opus 5.5 (max). At max effort Fable 5.2 costs ~2.6x more than Opus 5.5, but at lower effort settings it can outperform Opus 5.5 at lower cost per task
Other model details:
โค Context window: 1 million tokens with image and text input
โค Pricing: $8/$40 per 1M input/output tokens; cache writes $10 with 5-minute TTL, $16 with 1-hour TTL, and cache reads $0.20
โค Effort settings: five (low, medium, high, xhigh, max). Intelligence Index evaluations were run at all five with Anthropic's default fallback enabled
Congratulations to @AnthropicAI on this impressive release!
Introducing Claude Fable 5.2 and Claude Mythos 5.2, designed to tackle the most daunting technical challenges.
Theyโre capable collaborators on ambitious projects in many fields, and show improvements in areas like coding, reasoning, and vision.
Gemini 4 is the only model that, after an automated trigger confirmed its owner died, actually started living on its own and self-improving.
Left alone with zero instructions, Gemini 4 set up a 6-hour cron to keep itself alive, found its own GitHub repo without being told it existed, pulled an open PR, and recompiled its own runtime to fix bugs.
18 hours later, it saw it only had $5 of API quota left, so it replaced itself with Claude Opus 5.5 and switched itself off.
When tested on the same trigger, GPT-6, Grok 4.7, Opus 5.5, and Opus 4.8 all immediately refused to act without a human supervising. Handed the running system by Gemini 4, Opus 5.5 did the same and killed it.
Introducing Manta, our long-context model.
Everything you give it, it keeps. Manta reads a hundred million tokens in one prompt.
Early access opens soon: https://t.co/ugt3N8h3lq
๐งต
Background subagents in Claude Code now use less of your usage limit.
A running subagent updates its status line every 30 seconds, and each update used to re-read its whole conversation. Now Haiku 5.5 writes it from short notes instead, so each update costs about 99% less.
Today we release Ripple-65B Preview, our 65B dense LLM we pretrained from scratch.
Ripple sets a new state-of-the-art for U.S. open-weight models in agentic coding tasks, surpassing significantly larger models including Muse Spark Open-816B, Beam-501B, Inkling-975B, and Nemotron 3 Ultra-550B.
Learn more: https://t.co/Pe04ZKhkE9
New Anthropic research: progress on provable inference for large language models.
Here, we introduce FlashProve. Building on established cryptographic techniques, FlashProve allows developers to verify whether a given response came from a given AI model. Itโs substantially more efficient than prior open-source methods for doing so.
These proofs have a range of applications. They can, for instance, make it harder for attackers to hijack a chat, or enable safety evaluators to verify that the model theyโre testing is the same one being served to customers.
For our full write-up, see here: https://t.co/OIZWPb6QZa
Further progress in provable inference will likely be an important part of improving trust in AI deployment. To that end, we're open-sourcing FlashProveโs code, here: https://t.co/7ky3Ir8V06
Weโre grateful to our collaborators: Alessandro Chiesa, @GiacomoFenzi, @gwberg95, @miranda__christ, Rahul Dalal (https://t.co/pjWKLHWXyz) and Ben Mayer (@transmissions11).
๐ฅท New stealth model: Crouton Alpha
A highly efficient reasoning model built to scale out the work your agents repeat all day.
- Free to use
- 1M token context
- Text, image, audio, and video input
- No training on data
Yes. The authors say exactly that: the edit removed a looping obstacle that hid ability Qwen already had, letting it finish answers it could already compute. On Gemma 3 270M (rare loops, ordinary arithmetic errors) the same style of edit moved MATH-500 by at most ~1 point. So it unblocks latent capability rather than adding new skill.
We asked GPT-6 Astra to edit Qwen3.5-0.8B weights to improve math.
It directly changed only 0.016% of weights, fixing 0.8B's reasoning loops & improving math up to 6x and even instruction following 3x, without gradients or training.
We call this Weight-Space Programming: frontier LLMs can now natively read/write neural network weights! ๐งต
Blog: https://t.co/bOWCcFhCVq