AI

Harness Leaderboard - Kimi Code, Hermes, Pi, Claude Code, Codex, and Others Were Benchmarked Against Terminal Bench 2.1 & Deep SWE 1.1

Read full article at source ↗

Based on a feed excerpt; full article content has not been acquired. Source evidence retrieved 07/11/2026, 21:47:25 UTC.

Briefings

A Reddit thread from November 2026 reports benchmarking of AI coding assistants against terminal benchmarks, specifically Terminal Bench 2.1 and Deep SWE 1.1. The evidence indicates that Kimi Code, Hermes, Pi, Claude Code, and Codex were evaluated, though specific performance scores or comparative rankings are not provided in the source.

Key points

• AI coding assistants including Kimi Code, Hermes, Pi, Claude Code, and Codex were benchmarked against terminal benchmarks.
• The evaluation utilized Terminal Bench 2.1 and Deep SWE 1.1 as the primary test suites.

Context

This finding suggests a competitive landscape for AI coding tools, where models like Kimi Code and Hermes were tested against established terminal benchmarks to assess their coding capabilities.

Limitations

The source is a Reddit thread with limited scope; specific benchmark scores, comparative rankings, or detailed results are not provided in the evidence.

Generated by Bittrees-hosted qwen35-2b sha256:aaf42c8b7c3c on 08/11/2026, 00:56:10 UTC. Based on publisher-provided feed evidence; may not cover the full original article. Check the source for details.

View archived briefing on IPFS
bafybeih2nmoybktzdydqmhkhu7j34wmc6g5piygtci4c3ftj5huvvwz7qy