Skip to content
Hbench
hbench

Results

Results

Every published run, newest first, each with its own pass rate. Filter it, or download what you see as CSV.

56 runs

Download CSV

Select two rows of one board to compare them task by task.

Results, one row per published run
HarnessBoardPass rateHigher is better, 95% intervalRunCost per taskhbenchFinished
DeepSeek Harness0.2.1-rc.2
humaneval-plus@1.0 (unreviewed) on GPT-6.1 Sol96.3%93.3–98.8#37—0.3.624 Oct 2026
Codex0.164.0
humaneval-plus@1.0 (unreviewed) on codex-auto-review94.5%90.9–97.6#36—0.3.624 Oct 2026
Codex0.164.0
humaneval-plus@1.0 (unreviewed) on GPT-5.6 Terra95.7%92.7–98.2#35$0.030.3.624 Oct 2026
Codex0.164.0
humaneval-plus@1.0 (unreviewed) on GPT-5.6 Luna95.7%92.7–98.8#34$0.0030.3.624 Oct 2026
Codex0.164.0
humaneval-plus@1.0 (unreviewed) on GPT-6 Astra95.7%92.7–98.2#33$0.110.3.624 Oct 2026
Codex0.164.0
humaneval-plus@1.0 (unreviewed) on GPT-6 Luna88.4%83.5–93.3#32$0.0010.3.624 Oct 2026
Codex0.164.0
humaneval-plus@1.0 (unreviewed) on GPT-6 Sol95.1%91.5–98.2#31$0.030.3.624 Oct 2026
Codex0.164.0
humaneval-plus@1.0 (unreviewed) on GPT-6.1 Sol95.7%92.7–98.2#30$0.020.3.624 Oct 2026
Claude Code2.1.309
humaneval-plus@1.0 (unreviewed) on codex-auto-review93.3%89.0–97.0#23$0.100.3.624 Oct 2026
Claude Code2.1.309
humaneval-plus@1.0 (unreviewed) on GPT-5.6 Terra93.9%90.2–97.0#26$0.070.3.624 Oct 2026
Claude Code2.1.309
humaneval-plus@1.0 (unreviewed) on GPT-5.6 Luna93.9%90.2–97.6#24$0.100.3.624 Oct 2026
Claude Code2.1.309
humaneval-plus@1.0 (unreviewed) on GPT-5.6 Sol95.7%92.7–98.2#25$0.070.3.624 Oct 2026
Claude Code2.1.309
humaneval-plus@1.0 (unreviewed) on GPT-6 Astra96.3%93.3–98.8#28$0.060.3.624 Oct 2026
Claude Code2.1.309
humaneval-plus@1.0 (unreviewed) on GPT-6 Luna92.7%88.4–96.3#29$0.060.3.624 Oct 2026
Claude Code2.1.309
humaneval-plus@1.0 (unreviewed) on GPT-6 Sol95.1%91.5–98.2#22$0.070.3.624 Oct 2026
Claude Code2.1.309
humaneval-plus@1.0 (unreviewed) on GPT-6.1 Sol95.7%92.7–98.2#27$0.070.3.624 Oct 2026
Codex0.163.0
Terminal-Bench 2.0 on GPT-5.6 Sol62.9%52.8–73.0#21—0.3.523 Oct 2026
Claude Code2.1.302
SWE-bench Verified on Claude Opus 5.590.8%88.2–93.2#18$0.160.2.319 Oct 2026
DeepSeek Harness0.2.0-rc.2
SWE-bench Verified on Claude Opus 5.593.4%91.2–95.4#17—0.2.319 Oct 2026
Terminus 2
HumanEvalFix on Qwen2.5-Coder 7Bsample-40 task set, 1 attempt7.5%0.0–17.5#15—0.2.418 Oct 2026
clnkr0.3.11
Terminal-Bench 2.0 on GPT-5.5 (5 attempts, open network)66.1%57.5–74.2#89—converted7 May 2026
harness-agent
Terminal-Bench 2.0 on minimax-m2.7-highspeed (5 attempts, open network)42.5%33.9–50.8#95—converted2 May 2026
vixv0.1.22
Terminal-Bench 2.0 on claude-opus-4-7 (5 attempts, open network)89.9%84.7–93.9#127—converted1 May 2026
Capy
Terminal-Bench 2.0 on GPT-5.5 (5 attempts, open network)83.1%76.6–89.2#47—converted28 Apr 2026
little-coder0.1.24
Terminal-Bench 2.0 on qwen3.5-9b (5 attempts, open network)9.2%5.8–13.3#101—converted27 Apr 2026