jlehman_ adds decision model support to OpenClaw core and plugins
OpenClaw says the support was pushed last week, and describes how it is thinking about using decision models and how people can contribute.
AI News
LangSmith now supports decision models including Jev and SemIf
LangSmith provides visibility into every step for supported decision models, including Jev and SemIf.
Artificial Analysis compares text-to-speech models across 9 new languages
The multilingual Controlled Voice Arena leaderboards compare text-to-speech models in 9 new languages. A language-selectable leaderboard, a public voting arena, and the benchmarking methodology are available.
Four labs split open-weight leads on Artificial Analysis TTS language boards
No single open-weight text-to-speech model leads all nine languages in Artificial Analysis’s rankings. Higgs Audio V3 TTS, OpenAudio S1 Mini, Voxtral TTS, and Breeze TTS 2 each top the open-weight field in different languages, and each board includes only 2 to 6 open-weight models.
Artificial Analysis adds text-to-speech leaderboards for nine languages
The Controlled Voice Arena now ranks text-to-speech models in Japanese, Mandarin Chinese, Hindi, Spanish, German, French, Portuguese, Vietnamese, and Arabic. Cartesia’s Sonic family leads eight of the nine boards, while Inworld AI’s Realtime TTS-2 ranks first in Mandarin.
Pi releases version 0.87.1 so its library can use the latest models
The 0.87.1 release updates Pi’s library to use the latest models, with the release published on pi.dev.
Expedia will join Muse so AI agents can sort travel plans
Expedia says it is joining Muse. Soon, a personal AI agent told a destination will be able to work with Expedia to sort hotels and everything in between.
Artificial Analysis places StepAudio 3 ASR at 88x real-time transcription
Artificial Analysis reports a transcription speed factor of 88x real time for StepAudio 3 ASR, behind three faster systems, near ElevenLabs Scribe v2 and Gemini 3.5 Transcribe, and ahead of Fun-Realtime-ASR-preview.
StepFun releases StepAudio 3 ASR at 1.7% WER on the AA-WER Index
Artificial Analysis says StepFun’s StepAudio 3 ASR is available through the StepFun API for non-streaming transcription, scoring 1.7% WER on the AA-WER Index, tied with Alibaba’s Fun-Realtime-ASR-preview and down from 4.7% for StepAudio 2.5 ASR. It transcribes at 88x real time and costs $0.40 per hour.
Pydantic AI agents now run on Jev
Pydantic says the agents use TypeSafe AI’s Jev classifier to answer typed questions, returning each agent’s existing output model with one confidence score per field.
Expedia is joining Muse so AI agents can arrange hotels
Expedia says a personal AI agent on Muse will soon be able to take a destination and work with Expedia to arrange hotels and other parts of a trip.
GPT 6 Sol, Luna and Opus 5.5 are now available in OpenCode
OpenCode now offers GPT 6 Sol, Luna and Opus 5.5. No pricing, limits, or eligibility conditions are stated.
Arena.ai says scores for GPT-6 Sol and GPT-6 Luna are coming soon
OpenAI’s two models can be tested on Arena, where votes on real-world agentic tasks feed the leaderboard. Arena.ai says petergostev has also compared GPT-6 Sol with GPT-5.6 Sol using the same prompts at max reasoning.
Opus 5.5, GPT-6-Sol and GPT-6-Luna are live on Radius
Pi says the three models are available on Radius at radius.earendil.com.
GPT-6 Sol is now available in Perplexity and Computer
GPT-6 Sol can be used in Perplexity and Computer, where it is the default Light option in the effort selector.
Claude says Claude Opus 5.5 is available today
Claude says Claude Opus 5.5 is available today. The statement was posted at 18:55 UTC on September 22, 2026, and gives no platform, price, quota, or eligibility terms.
GPT-6 Sol and GPT-6 Luna can be tested in Battle Mode and Agent Mode
On Arena.ai, both models can be tried in Battle Mode and Agent Mode.
Arena.ai adds GPT-6 Sol and GPT-6 Luna to Agent Arena and Code Arena
OpenAI’s GPT-6 Sol and GPT-6 Luna can now be tested in Arena.ai’s Agent Arena, with scores still to come, and in Code Arena for WebDev, Text, Vision, Search, and Document. OpenAI says the models build on GPT-6 Astra and that their API prices are 50% lower than GPT-5.6 promotional pricing.
GPT-6 Sol and GPT-6 Luna are generally available in GitHub Copilot
GitHub Copilot now generally offers two additional OpenAIDevs GPT-6 models: GPT-6 Sol for interactive and agentic coding, and GPT-6 Luna, which GitHub calls a lightweight, lowest-cost option for smaller tasks.
PrimeScientist gets 10.3% more reward with 50.6% fewer attempts than AutoResearch
DAIR.AI describes PrimeScientist, which uses an adaptive MCTS policy to choose where a research agent spends its budget. On 12 AI research tasks under the same budget, it reports 10.3% more reward and 50.6% fewer research attempts than AutoResearch.