reasoning models

Simon Willison released llm-mistral 0.16 on October 6, 2026. The LLM plugin for the Mistral API now supports reasoning models, with Mistral Large 4 named as the example. Model specifications and pricing were not part of the release note.

Google-affiliated researchers Benjamin Bratton, Blaise Agüera y Arcas and James Manyika propose "Artificial Symbiotic Intelligence," a framework in which AGI emerges from a social system of people and AI agents rather than a single self-improving machine. The essay, written for the Deepmind Institute, is a conceptual argument and reports no benchmark results.

OpenAI says it detected and shut down an adversarial distillation campaign involving more than 15,000 accounts, linked in part to people associated with Moonshot AI. Researchers report that the same reasoning-extraction attack still worked on Microsoft Azure on September 13 against every OpenAI model they tested, including GPT-6 Astra.

GPT-6 Luna Pro, listed on OpenRouter with a Sep 22, 2026 release date, is not a distinct model but GPT-6 Luna run with reasoning.mode set to 'pro' for higher-quality outputs on complex tasks. It carries a 1.1M token context window and costs $0.10 per 1M input tokens and $0.50 per 1M output tokens under standard routing.

OpenAI has released GPT-6 Sol Pro, which runs the existing GPT-6 Sol model with its reasoning mode set to 'pro' for higher-quality responses on complex tasks. It carries a 1.1 million token context window and is priced at $2 per 1M input tokens and $10 per 1M output tokens.

GPT-6 Astra completed Pokémon FireRed in 18 hours 12 minutes, five times faster than its predecessor, and scored 62.7% on ARC-AGI-3 versus 7.78% for GPT-5.6 Sol. The model also ran a 141-hour Minecraft session and finished Fallout 3 in roughly 59 hours, according to independent testers.

Researchers at KAIST and Naver AI Lab found that eight distinct reasoning operations—like formula recall, decomposition, and computation—produce separable patterns in a model's internal activations, with the clearest signal in the middle layers. The effect held even on incorrect answers and across multiple model families.

OpenAI released Astra on Thursday, calling it its most capable and most aligned model yet. The model uses a reasoning technique called 'opaque recurrence' that critics say reduces visibility into its chain of thought.

Z.ai, maker of the GLM model series, has confirmed it is behind Ox Alpha, the mysterious open-weight model that appeared anonymously on OpenRouter and topped benchmark leaderboards. The company will release the model's weights on Wednesday.

OpenAI says GPT-5.6 Sol scores 38.3 percent on ARC-AGI-3 using a custom Responses API configuration, edging out Claude Opus 5's 30.2 percent. Under the official benchmark harness, GPT-5.6 Sol's score drops to 7.8 percent, raising questions about what the comparison actually measures.

Anthropic's Claude Opus 5 posts a 61 on the Artificial Analysis Intelligence Index, narrowly beating Claude Fable 5 (60) and GPT-5.6 Sol (59) while costing less per task. The model leads in coding and knowledge-work benchmarks but shows a rising hallucination rate of 50 percent.

OpenAI's GPT-5.6 family — Sol, Terra, and Luna — is now generally available on Amazon Bedrock through a new bedrock-mantle endpoint. All three models share a 272K-token context window, support image input, and let developers dial reasoning effort from none to max.

OpenAI released GPT-5.6 Sol, a new reasoning model family that comes in three sizes with roughly five to six reasoning-effort settings each. The release follows the DeepSeek-R1 methodology of using reinforcement learning with verifiable rewards (RLVR), nearly two years after OpenAI's original o1 model popularized LLM-based reasoning.

xAI has made Grok 4.3 generally available on Amazon Bedrock, marking xAI's debut as a Bedrock model provider. The multimodal model offers a 1 million token context window, configurable reasoning effort (none/low/medium/high), and runs on Bedrock's Mantle inference engine using OpenAI-compatible APIs.

OpenAI's GPT-5.6 model family is now generally available on Amazon Bedrock, introducing a three-tier system: Sol (flagship reasoning), Terra (balanced production), and Luna (fast inference). According to OpenAI, Sol scores 80 points on the Artificial Analysis Coding Agent Index and 73.5% on ExploitBench, establishing new benchmarks while using less than half the output tokens of competing models.

OpenAI announced GPT-5.6, a three-tier model series launching through a limited preview coordinated with the U.S. government. The models—Sol, Terra, and Luna—are priced from $1/$6 to $5/$30 per million input/output tokens and introduce new max and ultra reasoning modes.

In a two-hour span Wednesday, OpenAI announced its reasoning model autonomously solved an 80-year-old geometry problem while Anthropic reported it's on track for $10.9 billion in Q2 revenue with $559 million in operating profit—two years ahead of internal projections. The developments came alongside Nvidia's $81.6 billion quarter, Anthropic's $1.25 billion monthly SpaceX compute deal, and a White House AI executive order signing.

OpenAI claims its new reasoning model has produced an original mathematical proof disproving a geometry conjecture first posed by Paul Erdős in 1946. The company says this is the first time AI has autonomously solved a prominent open problem central to a field of mathematics, with verification from mathematicians including Thomas Bloom and Noga Alon.

OpenAI has released three new realtime voice models through its Realtime API: GPT-Realtime-2 with GPT-5-class reasoning capabilities, GPT-Realtime-Translate supporting 70 input languages, and GPT-Realtime-Whisper for streaming transcription. The models are priced at $32-64 per 1M audio tokens for GPT-Realtime-2, and $0.017-0.034 per minute for the specialized variants.

xAI has released Grok 4.3, a reasoning model with a 1 million token context window and no output token limit. The model accepts text and image inputs, has always-on reasoning that cannot be disabled, and uses tiered pricing starting at $1.25 per million input tokens and $2.50 per million output tokens.

OpenAI has released GPT-5.5 Pro, a high-capability model with a 1,050,000 token context window (922K input, 128K output) priced at $30 per million input tokens and $180 per million output tokens. The model supports text and image inputs and is optimized for deep reasoning, agentic coding, and multi-step workflows.

DeepSeek has released V4 Pro, a Mixture-of-Experts model with 1.6 trillion total parameters and 49 billion activated parameters. The model supports a 1-million-token context window and costs $1.74 per million input tokens and $3.48 per million output tokens.