Open models and agentic coding hit deployment scale this week
Mistral 3, Claude Code and GPT-5.1 Codex Max, and The Deploy Log for this week. Every price and claim read off the publisher's own page, with a call on each.
Mistral
Mistral 3
DEPLOYif you run open models in production
WHAT SHIPPEDMistral 3 ships three dense models at 14B, 8B, and 3B parameters plus Mistral Large 3, a sparse mixture-of-experts with 41B active and 675B total parameters, all under Apache 2.0. The 14B reasoning variant hits 85% on AIME '25.
THE CATCHThe Ministral instruct models match or exceed comparable models while often producing an order of magnitude fewer tokens, so cost per task can drop even when the model card looks similar.
IF YOU RUNedge or local workloads and want a model family with image understanding across base, instruct, and reasoning variants
SOURCEMistral, Introducing Mistral 3
Anthropic
Claude Code
DEPLOYif you build software with JavaScript or TypeScript
WHAT SHIPPEDClaude Code reached $1 billion in run-rate revenue in November, six months after becoming generally available in May 2025. Anthropic is acquiring Bun, the JavaScript runtime with more than 7 million monthly downloads and over 82,000 GitHub stars.
THE CATCHBun will remain open source and MIT-licensed, so the acquisition does not lock you into Anthropic's stack even as Claude Code gets faster performance and new capabilities.
IF YOU RUNa JavaScript or TypeScript codebase and want faster installs, bundling, and test runs inside agentic coding workflows
SOURCEAnthropic, Anthropic acquires Bun as Claude Code reaches $1B milestone
Vercel
GPT-5.1 Codex Max
DEPLOYif you use Vercel AI Gateway
WHAT SHIPPEDOpenAI's GPT-5.1 Codex Max is now available on Vercel AI Gateway with no other provider accounts required. The model uses compaction to operate across multiple context windows and is optimized for long-running coding tasks.
THE CATCHSet reasoningSummary to auto and reasoningEffort to low in providerOptions to get the token efficiency the model is trained for.
IF YOU RUNlong-running coding agents that need to maintain context and reasoning without starting new sessions
SOURCEVercel, GPT-5.1 Codex Max now available on Vercel AI Gateway
The Deploy Log
Models and APIs
| WATCHif you maintain a model integration Transformers v5. Transformers v5.0.0rc-0 launches with over 400 model architectures, more than 750,000 compatible checkpoints on the Hub, and 1.2 billion total installs, up from 20,000 installs per day at v4. Hugging Face |
| DEPLOYif you run Snowflake in a regulated industry Snowflake partnership. Anthropic and Snowflake announce a multi-year, $200 million agreement making Claude models available to more than 12,600 global customers across Amazon Bedrock, Google Cloud Vertex AI, and Microsoft Azure, with greater than 90% accuracy on complex text-to-SQL tasks. Anthropic |
| WATCHif you process text, images, and videos Nova 2 Lite. Amazon's Nova 2 Lite reasoning model is now available on Vercel AI Gateway with no other provider accounts required, processing text, images, and videos to generate text. Vercel |
| DEPLOYif you use Vercel AI Gateway Mistral Large 3. Mistral Large 3 is now available on Vercel AI Gateway with no other provider accounts required, using a sparse mixture-of-experts architecture with 41B active parameters and 675B total. Vercel |
| DEPLOYif you run agent-style tool calls DeepSeek V3.2. DeepSeek V3.2 and V3.2 Speciale are now available on Vercel AI Gateway with no other provider accounts required, with V3.2 supporting combined thinking and tool use in both reasoning and non-reasoning modes. Vercel |
| WATCHif you need an open weight reasoning model Trinity Mini. Arcee AI's Trinity Mini, an open weight MoE reasoning model with 26B parameters and 3B active, is now available on Vercel AI Gateway with no other provider accounts required. Vercel |
| WATCHif you run math reasoning workloads DeepMath. Intel's DeepMath agent, built on Qwen3-4B Thinking and fine-tuned with GRPO, reduces output lengths by up to 66% while improving accuracy on MATH500, AIME, HMMT, and HLE benchmarks. Hugging Face |
| DEPLOYif you work at a nonprofit Claude for Nonprofits. Anthropic launches Claude for Nonprofits with discounts of up to 75% on Team and Enterprise plans, connectors to Blackbaud, Candid, and Benevity, and a free AI Fluency for Nonprofits course. Anthropic |
Tools and products
| DEPLOYif you run Next.js 15.0.0 through 16.0.6 React2Shell protections. Vercel releases npx fix-react2shell-next to update affected Next.js apps, with Vercel Agent able to open pull requests that upgrade vulnerable projects and WAF rules filtering known exploit patterns. Vercel |
Left out
- Next.js vulnerability blocking: a figure in the draft could not be found on the publisher's own page
- Deploy step speed: a figure in the draft could not be found on the publisher's own page
- Runtime logs speed: a figure in the draft could not be found on the publisher's own page
- HF Skills training: the publisher's page does not show it available yet
- NORAD Tracks Santa: the publisher's page does not show it available yet
- AWS databases: the publisher's page does not show it available yet
- swift-huggingface: it did not carry a number that moved
- Vercel Agent installs: it did not carry a number that moved
- Remend: it did not carry a number that moved
- Runtime logs rewrites: it did not carry a number that moved
- Notion Agent forms: it did not carry a number that moved
- Vercel for Platforms: it did not carry a number that moved
- Platform Elements: it did not carry a number that moved
The pattern
This week the deployment surface widened in two directions at once: open model families got cheaper and more capable at the edge, while agentic coding tools crossed from experiment to revenue at scale. Mistral 3 puts an 85% AIME '25 reasoning model under Apache 2.0, and Claude Code hit $1 billion in run-rate revenue six months after general availability.
The infrastructure underneath is consolidating around that shift. Vercel now routes GPT-5.1 Codex Max, Mistral Large 3, DeepSeek V3.2, Nova 2 Lite, and Trinity Mini through one gateway with no separate provider accounts, while Anthropic bought Bun to speed up the JavaScript toolchain that agentic coding depends on. The practical question for a team is no longer which model to try, but which gateway and runtime to standardize on.