LLM release radar

GPT-6.1 Terra launches in ChatGPT (Work), Codex and the API at $1/$5 per 1M tokens - half of GPT-5.6 Terra - for fast, cheap computer use; OpenAI says it matches GPT-6 Astra at ~1/10 the token price.

428B-total / 23B-active weights published on Hugging Face (MXFP8) under the MiniMax community license.

Kimi K3.1 Pro (2.8T total / 104B active MoE, 1M context, image input) launches in the Kimi apps and API at $2/$10 per 1M tokens, plus a 2-3x faster Highspeed tier; weights promised by Nov 4. AA Intelligence Index 55.

M3.1 Flash moves from Token-Plan preview to public launch at $0.15/$0.60 per 1M tokens.

Mercury 3, Inception's flagship diffusion LLM (1M context, ~1,000 tok/s claimed), launches at $0.20/$0.70 per 1M tokens and replaces Mercury 2/2.5.

Mercury 2 and 2.5 closed to new users at the Mercury 3 launch; existing customers keep access until 2026-11-15.

Updated V4.1-Flash checkpoint with stronger reasoning, coding, agent and visual understanding and shorter outputs; MIT weights on Hugging Face.

Beam-1.0-501B-A23B weights (BF16 / FP8 / NVFP4) published on Hugging Face under Apache-2.0.

Ling-3.1-flash (560B / 25B active MoE) moves from trial to full release with MIT-licensed fp8 weights.

74B-A9B 2-bit (~24 GB) local-agent model derived from Super VL; OpenMDW-1.1 weights.

Open 120B-A12B hybrid Mamba-Transformer vision-language MoE for multimodal agents, OpenMDW-1.1 weights.

Meta releases Muse Spark Open (816B total / 42B active MoE, 1M context, multimodal input) with downloadable weights under the Muse Model License.

Gemini 4 Argon (gemini-4-argon) goes stable in the Gemini API and app; $2/$10 per 1M through Nov 30, then $4/$20.

24B-total / 2.45B-active MoE with 131K context, open weights under LFM Open License v1.0; $0.10/$1.00 on Liquid Console.

Larger Muse Wonder Full ships to Muse Code Power/Maximum subscribers only; AA Index 56.

Muse Wonder enters API preview at $4/$20 per 1M tokens with 1M context and multimodal input.

JetBrains released Mellum2.1 (Mellum2.1-12B-A2.5B-Thinking) on 2026-10-08: an open-weight sparse Mixture-of-Experts code model, 12B total / 2.5B active (28 layers, 64 experts, 8 active/token), 131,072-token context, text + code. A reasoning ("thinking") model post-trained from the Mellum2 base with reinforcement learning in real repository sandboxes, carrying the Mellum2 architecture unchanged. Apache-2.0 weights on Hugging Face (Transformers / vLLM / SGLang + GGUF for llama.cpp, Ollama, LM Studio); built for private self-hosting, no first-party hosted API. JetBrains reports large agentic-coding gains over Mellum2 - SWE-bench Verified 2.0 -> 47.0, SWE-bench Pro 0.0 -> 28.0, Terminal-Bench 2.1 0.6 -> 17.4 - plus LiveCodeBench v6 82.0, HumanEval+ 91.5, AIME 25/26 83.3 and GPQA Diamond 64.6 (vendor-reported).

Anthropic released Claude Haiku 5.5 on 2026-10-07, its cheapest and fastest small model to date, succeeding Claude Haiku 4.5. It is the first Haiku-class model with an adjustable effort control (adaptive thinking on by default, effort default medium). Text+image in / text out, 1M-token context, 128K max output (up to 300K via Batch API beta), June 2026 knowledge cutoff. Proprietary API-only; GA via the Claude API (claude-haiku-5-5), Amazon Bedrock, Google Cloud Vertex AI, Microsoft Foundry and the Claude Platform on AWS. Tiered pricing: $0.10/$0.50 per 1M in/out up to 100k tokens, $0.50/$2.50 above 100k ($0.01/1M cache reads, $0.125/1M cache writes, 50% off via batch). Benchmarks: OSWorld 2.1 72.4%, Terminal-Bench 4.0 39.2%, FrontierCode 1.1 46.4%, Humanity's Last Exam 45.9% (57.4% with tools), GDPval-AA v2.1 1620.

Mistral AI launched Mistral Large 4 ("Le Chonk") in public preview on 2026-10-06: a granular Mixture-of-Experts flagship (first-party docs: 1.05T total / 52B active; count disputed — some press 49B active, one analysis cites 675B/41B), natively multimodal (text + image via a ~1.6B vision encoder) with text output, 160+ languages, and a 1M-token context, trained on 3,800 NVIDIA GB200 GPUs in European datacenters. Available via the Mistral Studio preview API at $1.36 / $4.18 per 1M in/out tokens ($0.14 cached); open weights promised by end of October 2026 (license not yet announced). Vendor benchmarks: Cybench 93%, CyberGym-E2E 82%, DeepSWE v1.1 61.7%, AA coding-agents 49.8%, Lakera B3 93.3%.

Reflection AI debuted Beam on 2026-10-05: a sparse Mixture-of-Experts text LLM with 501B total / 23B active parameters (52 layers), trained on ~23.8T tokens, with a 1M-token pretrained context (API beta capped at 262,144). Apache-2.0 open weights are promised later in October 2026; until then access is via an OpenAI-compatible API beta (model id Beam-501B-A23B). Reflection reports SWE-bench Verified 80.9, Terminal-Bench v2.1 80.1, DeepSWE v1.1 44.4 and Humanity's Last Exam 36.2, claiming parity with GLM-5.2 on reasoning at 3-4x less inference compute. Vendor-reported, unverified at launch.

Aleph Alpha, the German sovereign-AI company, released Kolibri-1 on 2026-10-03 under the Apache-2.0 license with open weights on Hugging Face (an FP8 flagship, a BF16 base and several quantized variants). A Mixture-of-Experts text model with 78.1B total / ~3.46B active parameters (50 layers, 384 routed experts, 6 active per token), tuned for bilingual German/English use and trained natively to a 262,144-token context, extensible to 1,048,576 (1M). Text in / text out with training for grounding, tool-calling and abstention; built and trained in Germany and Finland. Reported post-training scores: 75.5% overall English, 70.8% German, 96.5% math and 89.3% code (English averages), and 63.2% on RULER at 1M context. Knowledge cutoff 2026-06-18; figures vendor/self-reported.

Bilibili's Index LLM team open-sourced the Index-Translate family on 2026-09-30 under Apache-2.0, led by Index-Translate-35B-A3B-preview — a Qwen3.5-based Mixture-of-Experts translation model (35B total / ~3B active, 262,144-token context) covering 150 languages. Weights are on Hugging Face (IndexTeam) and ModelScope alongside a technical report and online demo, with 2B and 9B text siblings plus speech-translation, dubbing and long-document companion models. Reported results include FLORES COMET-22 0.8794, WMT26 Judge 76.76/100 and instruction-following IFscore 0.8336. A narrow translation specialist and Qwen3.5 derivative; figures vendor/self-reported.

Google DeepMind unveiled Gemini 4 Argon on 2026-09-30, billed as its most powerful model yet and the first new frontier Gemini since Gemini 3.1 Pro. Text/image/video in, text out; 1,048,576-token (1M) context and an expanded output budget (Google cites up to ~1M output tokens vs the prior 64K). Reasoning model with extended thinking; proprietary, weights not published. Staged rollout — Fairwind Program (trusted cyber defenders) first, then paid API customers and Google AI Ultra. Introductory pricing $2/$10 per 1M in/out ($0.10 cached), reverting to $4/$20 after the intro period. Google-reported leaderboard leads: DeepSWE 1.1 77.9%, FrontierSWE V2 55%, LAB-Bench 2 88.8%, LVBench 91.7%; Artificial Analysis Intelligence Index 53. Figures vendor/third-party-reported.

Ant Group's inclusionAI lab launched Ling-3.1-flash in late September 2026 (2026-09-29/30), the successor to Ling-3.0-flash. A 560B-total / ~25B-active Mixture-of-Experts reasoning model with an explicit thinking mode for agent tasks, search, office software, and specialist apps. Text in/out; 262,144-token (256K) context in the launch trial (1M target once enabled), up to 32,768 output tokens. Offered as a two-week free trial via Vercel AI Gateway (hosted by Novita); inclusionAI says open weights will follow after the trial, so weights/license/price are announced but not yet posted. inclusionAI-tracked benchmarks: aggregate agentic 81.0 (5 benchmarks), Terminal-Bench 4 40.4%, SWE-Atlas 55.9%, HealthBench Professional 65.3%. Figures vendor/third-party-reported.

OpenAI released GPT-6.1 Sol on 2026-09-29 at DevDay 2026, an agentic-work upgrade to GPT-6 Sol. API id gpt-6.1-sol; text+image in / text out. OpenAI reports it nearly matches GPT-6 Astra's intelligence on agentic coding, computer use, and professional work at about one-fifth of Astra's token prices. 1.05M context (maximum 922K input tokens), 128K max output, knowledge cutoff 30 April 2026. Pricing $2/$10 per 1M in/out ($0.10 cached input, $2.50 cache writes); prompts over 272K input tokens bill at 2x input/cache and 1.5x output. Available via the API and in ChatGPT Work and Codex for Plus, Pro, Business, Enterprise, and Edu users.

OrcaRouter (Continuum AI, San Francisco) released OrcaSAQ-2 27B on 2026-09-28 under Apache 2.0 — its first published open-weights model and a heavily compressed build of Alibaba's Qwen3.8-27B via a proprietary sensitivity-aware mixed-precision quantization ("SAQ") that stores weights at ~3.21 bits, shrinking the 54 GB BF16 checkpoint to 12.3 GB so the 27.8B-parameter model serves on a single 16 GB GPU. Text-only (vision tower dropped), it keeps the base's 262,144-token context and hybrid-attention design and reports near-lossless fidelity (WikiText-2 perplexity 5.6482 vs 5.6468 BF16, 93.2% token agreement, 0.031 KLD) plus SWE-bench Verified 70.0% and Terminal-Bench 2.1 58.4%; ~65 tok/s single-stream (~90 with MTP speculative decoding), 330+ tok/s at 8-16 streams, served via vLLM. Free weights on Hugging Face; OrcaSAQ2 kernel on GitHub. First tracked model from OrcaRouter. Self-reported figures.

Anthropic released Claude Sonnet 5.5 on 2026-09-28, succeeding Claude Sonnet 5. It runs 30%+ faster and can cut the total cost of a task by up to 30% through faster output and fewer tool calls, at unchanged list pricing of $2/$10 per 1M input/output ($0.20/1M cache reads, $2.50/1M cache writes). Text+image in / text out, 1M-token context, 128K max output, June 2026 knowledge cutoff. It powers the Claude.ai free tier and is available via the Claude API (claude-sonnet-5-5), AWS Bedrock, Google Cloud Vertex AI, and Microsoft Azure. Benchmarks include Terminal-Bench 4.0 70.6%, CursorBench 4.0 55.5%, OSWorld 2.1 80.1%, and GDPval-AA v2.1 1844.

NaiveAI, a Beijing-based lab, released Naive-N0.5-Flash on 2026-09-27 under the MIT license with open weights and inference code on Hugging Face. A 309B-total / ~15.5B-active Mixture-of-Experts model continually pre-trained (~3.25T tokens) from Xiaomi's MiMo-V2.5, notable for an attention stack with no full/global-attention layers (39 sliding-window + 9 DeepSeek Sparse Attention layers across 48 layers, GQA-4) and a native 1,048,576-token (1M) context. Text in / text out for coding and AI-R&D work, with an Ultrafast serving mode cited up to ~2,000 tok/s. Hosted API launched invitation-only at $0.10/$0.40 per 1M in/out ($0.01 cache read). Vendor benchmarks span coding (SWE-Bench Pro, DeepSWE v1.1, Terminal-Bench 2.1) and AI-R&D (PostTrainBench, MLE-bench-30, PaperBench), published as figures only. A derivative/continued-pretrain of MiMo-V2.5; figures vendor/self-reported.

MiniMax quietly launched MiniMax-M3.1-Flash-Preview on 2026-09-27 inside the MiniMax Code tool and via its Token Plan subscription, with OpenAI- and Anthropic-compatible API endpoints (id MiniMax-M3.1-Flash-Preview). A fast, lower-cost coding model for everyday development work (bug fixes, small features, routine tasks) with always-on tunable reasoning across five effort levels. Text/image/video in, text out; 1,000,000-token context. Closed weights, no public model card or parameter count at launch. MiniMax list pricing $0.30/$1.20 per 1M in/out (CometAPI resells ~$0.24/$0.96). Preview release with no first-party benchmark table published at launch.