It’s about twice as expensive as Opus, but Fable 5 / Mythos 5 is the most capable AI on the planet (Also, why does every model seem to suck on the legal agent benchmark?)
AGENTS
-
Fable 5 available in Claude Code and Cowork, best coding model
By
–
Fable 5 is now available in Claude Code and Cowork
— Boris Cherny (@bcherny) 9 juin 2026
Fable is the best model I have used for coding, by a wide margin. It is a big step up, enabling less prompts and steers, more efficient token use, better code quality, better tool use, more intelligent self-verification, longer… https://t.co/RmVfZh39HtFable 5 is now available in Claude Code and Cowork. Fable is the best model I have used for coding, by far. It is a big step forward, allowing fewer instructions and guidance, more efficient token usage, better quality of
-
Improvement of vision models on Pokémon Red
By
–
The model also improves on vision tasks. Very interesting what they comment: where previous models could not surpass Pokémon Red in an agentic way, even by supplementing their vision with additional information and tools,
-
Big leap in agentic programming and new FrontierCode Diamond benchmark
By
–
Regarding the first benchmark table they share, the big leap is observed especially in agentic programming (where labs get more value from users). The new FrontierCode Diamond benchmark released yesterday, designed to be more difficult than
-
Snorkel AI joins Agents’ Last Exam, forecasts agents for all jobs by 2027
By
–
We're proud Snorkel AI is part of Agents' Last Exam, with our researchers @amanda_dsouza and @vincentsunnchen among the co-authors and support from our Open Benchmarks Grants initiative.
— Snorkel AI (@SnorkelAI) 9 juin 2026
The forecast: agents will do almost every job by 2027. The result on real, code-graded work?… https://t.co/QmhOGdTNQ9 pic.twitter.com/FSnsszIszOWe're proud Snorkel AI is part of Agents' Last Exam, with our researchers @amanda_dsouza and @vincentsunnchen among the co-authors and support from our Open Benchmarks Grants initiative. The forecast: agents will do almost every job by 2027. The result on real, code-graded work?
-
Consensus on answers hides deeper reasoning disagreement in agent debates
By
–
// The Consistency Illusion // Multi-agent debate can make agents agree on the final answer while their underlying reasoning stays misaligned. This work finds that consensus on the output hides disagreement on the path that produced it, and you only ever see the output. A lot
-
Fable’s jump in capability: works 9+ hours on 15-page design
By
–
I've had access to Fable for a bit. A genuine jump in capability, I could feed it a 15 page design document for a project and it would work for 9+ hours and deliver terrific results. But working with it is weird & weirder is coming Lots of examples:
-
Free LangChain Academy course to monitor and improve agents
By
–
Sign up for free to our LangChain Academy course and get ready to learn how to monitor + improve agents in production.
-
AI garbage flood pushing open-source developers to limit
By
–
Flood of #AI 'garbage' is pushing open-source developers to the limit
by Matthew Sparkes @newscientist Learn more: https://
bit.ly/4vofwaU #Coding #GenAI #AIAgents #ArtificialIntelligence #MachineLearning -
Evaluation of interaction and spatial reasoning of multimodal agents
By
–
SpatialWorld Evaluation of interaction and spatial reasoning of multimodal agents in real-world tasks