Data and AI releases

548 this month · 176 companies

open-instruct 0.3.0

UpdateNot yet verified

This is a big one, as I haven't released since March. Whoops! Also my last release as maintainer, as today is my last day at Ai2. I'll miss you all. Added - Add minimax-m3 to PRICE_PER_MILLION_TOKENS in open_instruct/judge_utils.py for cost tracking on the latest MiniMax flagship model.

PricingObservability

Read on github.com

Also shipped Jun 11, 2026Ai2 in June 2026 →

Coding sessions in Linear

Earlier this year, we launched Linear Agent, giving teams a new way to plan and coordinate their issues and projects. Since then, we've added Code Intelligence to extend that understanding to the codebase, and MCP support to bring all of your context together. Today we're closing the loop: Linear Agent can now write code using Claude Code and Codex.

Preview
Source ↗

Import apps using MCP

Live on cloud • All plans Public beta 4.0 Edge • All plans Public beta 4.0.17 Stable • All plans Public beta You can now use Retool's MCP server to import React apps directly from your agentic coding environment. Open a React app, and ask Retool to import the app.

Beta
Source ↗

New External Guides and Surveys MCP tools

Description: AI assistants using the Amplitude MCP server now have a much richer view of your Guides and Surveys configuration. Previously limited to basic retrieval, the external MCP server can now fully understand a guide's published state, prioritization, and all other Guides and Surveys concepts , so AI-assisted workflows can reason about and act on...

GA
Source ↗

More from open-instructEvery open-instruct release →

open-instruct 0.2.0

Notable changelog items: 1. Fixed a major but where ZeRO-2 was discarding gradients. 2. We now support training with a RLEnvironment abstraction, including code execution sandboxing. 3. The Olmo-core DPO is ready to go, with support for TP, packing, and torch.compile. This should be much faster than the previous HF implementation. 4.

Update
Source ↗

open-instruct 0.1.0

We have made the following changes to open-instruct: Added - Added OLMo-core based DPO training script (). - Added SLURM scripts for OLMo SFT training with checkpoint resume support and configurable shuffle seed. - Added retry logic with exponential backoff to make_api_request for tool API calls (retries on timeouts, connection errors, 429, and 5xx).

Update
Source ↗