The latest Model releases stories across our sources, prepared from the publishers’ own reporting.
Mistral AI released Shieldstral, a 3B open-weights multimodal safety classifier under Apache 2.0. It accepts plain-language policies at inference time without retraining and outperforms open guard models up to 7x its size, according to the company.
A roundup of recent open model releases covers Thinking Machines' Inkling, Poolside's Laguna S2.1 and Laguna-XS-2.1, Tencent's Hy3, DeepSeek-V4-Flash-0731, Kimi K3, Meituan's LongCat-2.0, and Motif-3-Beta, among others.
Together AI announced a strategic partnership with Moonshot AI to natively serve Kimi models, beginning with Kimi K3, a 2.8T parameter sparse Mixture-of-Experts model with native vision and a 1M token context window. Together AI becomes a launch platform for Moonshot's open weights releases, offering day zero access and post-training.
Moonshot AI released Kimi K3, a 2.8T parameter MoE flagship model, on July 16th, with weights promised for July 27th. The model ranks high on several evaluation indexes, and the author reflects on open-weight competition between Chinese and American labs.
Mistral AI has introduced Robostral Navigate, an 8B model for embodied robot navigation that uses only a single RGB camera and no depth sensors or LiDAR. It reports 76.6% success on the R2R-CE validation unseen benchmark, outperforming multi-sensor approaches.
Mistral AI released Leanstral 1.5, a free Apache-2.0 licensed model with 119B total and 6B active parameters, reporting benchmark gains in formal verification in Lean 4, including saturating miniF2F and solving 587 of 672 PutnamBench problems.
Mistral AI released Mistral OCR 4, a document parsing model supporting 170 languages with bounding boxes, block classification and inline confidence scores. It runs in a single container for self-hosted deployments and is available via API at $4 per 1,000 pages, with Document AI priced at $5 per 1,000 pages.
Z.ai released its open-weight GLM-5.2 model to GLM Coding Plan members on June 13, 2026, with MIT-licensed weights and a release blog following on June 16. Commentators, including Interconnects' Nathan Lambert, described it as the first open model to feel right as a general agent in coding harnesses.
Together AI compared Kimi K2.7 Code and Claude Fable 5 on 12 generated landing pages, reporting Kimi cost about 94% less on average while scoring within a few points on nearly every page. A custom design-inspiration MCP server improved Kimi's output quality.
Ethan Mollick reports early access testing of Claude 5 Fable, the first publicly released Mythos-class AI model, saying it outperformed every public model he has used and worked up to a dozen hours on multi-page specifications. He describes feeling less like a director and more like a patron commissioning work he cannot watch being made.
NVIDIA Nemotron 3 Ultra, a 550 billion parameter open model with 55B active, is now available on Ollama's cloud for long-running agentic workflows. It offers 1M token context and is optimized for NVIDIA's NVFP4 format, with Ollama citing up to 30% cost savings versus other leading open models.
Mistral AI released Mistral Medium 3.5 in public preview, a 128B open-weight model that powers remote coding agents in Mistral Vibe and a new Work mode in Le Chat. It is available on Pro, Team, and Enterprise plans, priced at $1.5 per million input tokens and $7.5 per million output tokens.
Ethan Mollick reports early access to OpenAI's GPT-5.5, which he says outperforms prior models on a coding task and is faster than GPT-5.4 Pro. He also describes advances across OpenAI's apps and an updated image model.
Mistral AI released Voxtral TTS, a 4B-parameter open-weights text-to-speech model supporting 9 languages with low latency and voice adaptation. It is available via API at $0.016 per 1k characters and in Mistral Studio.
Mistral AI announced Mistral Small 4, an Apache 2.0 licensed model that unifies its Magistral reasoning, Pixtral multimodal, and Devstral agentic coding capabilities into a single model. It has 119B total parameters with 6B active per token, a 256k context window, and configurable reasoning effort.
Mistral AI released Leanstral, an open-source code agent for the Lean 4 proof assistant, with weights under an Apache 2.0 license, a free API endpoint, and integration into Mistral Vibe. Mistral reports Leanstral-120B-A6B outperforms much larger open-source models on its new FLTEval benchmark at lower cost.
Mistral AI released Voxtral Transcribe 2, two speech-to-text models: Voxtral Mini Transcribe V2 for batch transcription and Voxtral Realtime for live applications, which ships under the Apache 2.0 license. An audio playground in Mistral Studio was also launched.
Mistral AI released Mistral OCR 3, a document extraction model it says achieves a 74% overall win rate over Mistral OCR 2 on forms, scanned documents, complex tables and handwriting. It is available via API and a new Document AI Playground in Mistral AI Studio, priced at $2 per 1,000 pages with a 50% Batch-API discount.
Mistral AI released Devstral 2, a 123B coding model scoring 72.2% on SWE-bench Verified, plus Devstral Small 2 (24B, 68.0%), both open-source. It also introduced Mistral Vibe CLI, an open-source terminal coding agent. Devstral 2 is free via API before moving to paid pricing.
Ollama is partnering with OpenAI and ROOST to release the gpt-oss-safeguard reasoning models for safety classification, available in 20B and 120B sizes under the Apache 2.0 license.