Stories

Completed digest stories linked to this service.

• Claude Mythos posts record SWE-bench numbers, but it’s gated; tighten your evals...

  2026-04-08

  Anthropic’s Claude Mythos preview claims record SWE-bench results, but it isn’t publicly available and public ...
• OpenRouter’s coding leaderboard: free Qwen 3.6 Plus tops usage with 1M context a...

  2026-04-06

  OpenRouter’s latest usage data shows Qwen 3.6 Plus (free) leading coding workloads, with big context, solid re...
• Coding LLMs, March 2026: default to Sonnet 4.6, escalate to GPT-5.4, watch scaff...

  2026-03-22

  March 2026 coding LLM benchmarks show mid-tier models rival flagships, but scaffolding and cost drive real-wor...
• New long-horizon benchmarks say coding agents regress under maintenance; treat t...

  2026-03-11

  A new wave of long-horizon benchmarks shows most coding agents ship regressions over time, not just fixes. A ...
• Agents ace one-shot coding, but most break your code over months—time to harden ...

  2026-03-10

  New results say most coding agents cause regressions during long-term CI, and a new MassGen release adds built...
• MiniMax-M2.5 launches with SOTA coding claims; verify SWE-bench results

  2026-03-04

  MiniMax launched MiniMax-M2.5, a fast, low-cost coding and agentic model, but teams should validate its headli...
• Coding Benchmarks Shake-up: Qwen 3.5, MiniMax M2.5, and a SWE-bench Reality Chec...

  2026-03-03

  Open models like Alibaba’s Qwen 3.5 and MiniMax M2.5 post strong coding-agent results, but OpenAI’s audit of S...
• E2E agentic benchmarks replace SWE-bench; Gemini 3.1 favors deliberation

  2026-02-24

  Agentic coding benchmarks are shifting toward end-to-end app-building tests as SWE-bench Verified is being pha...