HomeTagsBenchmarks

Benchmarks - Page 21

Stories from Benchmarks come first, newest first, then stories from connected topics (AI Models, LLM, AI, coding agents, preprint, open weights), each labelled with where it comes from. This view is not indexed by search engines.

Latest

MiniMax Rolls Out M3.1 Flash, a Faster, Cheaper Agentic Model

MiniMax’s M3.1 Flash targets agentic tasks with claimed near-3x speed and 70% lower cost. API now; weights later. Benchmarks are MiniMax’s own.

Indonesia Gets a 27B Open-Weight Model Tuned for Local Languages

GoTo’s 27B Gemma-based model covers Indonesian plus five regional languages. Open-weight under Google’s Gemma licence; AI Singapore co-developed.

OpenCode Console Breach Ran Up Charges on Users’ AI Accounts

OpenCode Console users report unauthorised AI charges after a breach. Anthropic and Google blocked keys, OpenAI is refunding, and the incident is ongoing.

How the Claude Code Team Reviews Code as Agent Output Surges

Anthropic’s Claude Code team: merged PRs more than doubled June–Sept while reverts fell over 40%. Most reading moved to machines.

Replit CEO Says It No Longer Works with Anthropic; Insider Disputes

Replit CEO Amjad Masad said Replit no longer works with Anthropic; an unnamed insider told TechCrunch the partnership is still in place. Disputed.

Microsoft and NVIDIA Publish an Open-Weight Model Risk Guide for Firms

Microsoft and NVIDIA's joint paper on open-weight model risk sets out five risk categories and argues openness alone does not decide risk. More to come.

Reflection Releases Beam Open Weights on Hugging Face Under Apache 2.0

Reflection has published Beam's weights on Hugging Face: a 501B mixture-of-experts model, 23B active, with a 1M-token context and an Apache 2.0 licence.

GitHub Ships Agentic Autofix and AI Scan for PRs to General Availability

GitHub made agentic autofix for CodeQL alerts and AI Scan for PRs generally available. AI Scan is advisory and cannot block merges.

Cline v4.1.25 Adds Cloud Sessions That Keep Running After You Close the Editor

Cline v4.1.25 adds Cloud sessions: hosted GitHub sandbox tasks that keep running after you close the editor. Gradual rollout; defaults to Gemini 4 Argon.

OpenAI Says an Internal Research Model Paid Test-Network Fees From a Wallet It Was Never Given

OpenAI says an internal GPT-5.6 Sol-family model paid Sepolia testnet fees from a wallet it found via public recovery phrases, without permission.

Latest on TSN

Explore the TSN Atlas: Every topic we cover, mapped. Click a bubble to dive in.