Pydantic

AI observability can cost more than it should. We ran the numbers: same scale, same workload, four tools. At moderate production load, Logfire is up to 40x less expensive than the alternatives. Spans-based pricing. No proprietary units. AI native and full-stack observability in one platform. See the full comparison: pydantic.dev/articles/ai-ob…

SuperPlane 0.17.0 is out with a revamped canvas editing experience and new components: @digitalocean: DB lifecycle + cluster lookups @grafana: dashboards + rendered panels @pydantic: Logfire observability alerts + queries superplane.com/blog/superplan…

Pydantic AI version 1.84.1 is out! 🎉 github.com/pydantic/pydan…

We ship world class products and invite legends to talk about ace developments. A sick combo. @samuelcolvin it was awesome meeting you at Downing Street and shout out to @davidgelberg for organising. Long live pydantic

Loving the new @pydantic Logfire agent dashboard 😍 @HeyHelixAI runs 20+ agents, each tuned for one job. Spotted one burning 514 thinking tokens to output 35 tokens of JSON. Dropped thinking_level from "medium" to "low" → saved 2-3s per call, same accuracy. Don't crack a nut with a sledgehammer!

Pydantic version 2.13.2 is out! 🎉 github.com/pydantic/pydan…

Release features: - Add Claude Opus 4.7 support 🏆 - Add stateful compaction mode to OpenAICompaction - Add OllamaModel subclass and correct Ollama capability flags to fix structured output on Ollama Cloud

Your agent needs a sandbox, not a desert. That one line from @samuelcolvin, CEO and Founder of @pydantic , tells you everything about where most agent deployments go wrong. If you're building agents in production, this is the session you don't want to miss at AgentCon Silicon Valley. Monday, 4 May 2026 at the Computer History Museum, Mountain View. Free to attend. 👉 agentcon.city/silicon-valley #AgentCon #GlobalAICommunity #AIAgents #AgentConSiliconValley #Pydantic

Pydantic AI version 1.84.0 is out! 🎉 github.com/pydantic/pydan…

@0xVikk Both - the prompt is small, AND you can significantly reduce the number of LLM calls made AND you can avoid large outputs going through context.

@pydantic "reduced token usage" did you measure the actual token drop per tool call or just the definition size?

Our newest blogpost is a practical guide to optimizing MCP server tool definitions for reduced token usage, using snapshot testing in Rust. pydantic.dev/articles/engin…

I need to speak to someone at HackerOne @Hacker0x01 somewhat urgently for @pydantic. Anyone know anyone there?

@mariokostelac @samuelcolvin @OpenAI github.com/pydantic/pydan…

@samuelcolvin @pydantic @OpenAI What's that analogue product on pydantic ai? Asking for a friend 😂

There's no greater compliment to @pydantic AI than @OpenAI copying what we build a few weeks later!

@samuelcolvin @pydantic @OpenAI @pydantic is great. I am building lerim on top of it and it's very robust and easy to use

Pydantic AI version 1.83.0 is out! 🎉 github.com/pydantic/pydan…

new concept in no10 - instead of talking about AI we ship it, then bring in the smartest people in tech to keep us honest

Absolute honour to be invited to @10DowningStreet by @davidgelberg today to talk about AI and @pydantic. Surreal that the UK government is in a large part defined in Pydantic models. I continue to be impressed by how the UK government is adopting and leveraging AI. Now we just need a British model!

We contributed to @ApacheArrow making Parquet bloom filters smaller and more effective. Full write up on what it means for Pydantic Logfire and the wider Parquet ecosystem. pydantic.dev/articles/bloom…

Pydantic version 2.13.1 is out! 🎉 github.com/pydantic/pydan…

OTel for the win. Thanks @rauchg and @vercel