LLM Benchmarks — Scores, Methodology & Live Rankings
What each benchmark measures, how models score, and the current rankings
Claw-Eval Pass^3
Agentic Coding
Claw-Eval Pass^3 - third-pass pass rate on the Claw-Eval agentic coding benchmark.
SkillsBench Avg5
Agentic Coding
SkillsBench Avg5 - evaluated via OpenCode on 78 self-contained tasks (excluding API-dependent tasks); avg of 5 runs.
NL2Repo
Agentic Coding
NL2Repo - repository generation from natural language. Others evaluated via Claude Code (temp=1.0, top_p=0.95, max_turns=900).
QwenWebBench
Agentic Coding
QwenWebBench - internal front-end code generation benchmark; bilingual (EN/CN), 7 categories (Web Design, Web Apps, Games, SVG, Data Visualization, Animation...
SWE-bench Multilingual
Agentic Coding
SWE-bench Multilingual extends SWE-bench to multiple programming languages beyond Python.
Terminal-Bench 2.0
Agentic Coding
Terminal-Bench 2.0 - agentic coding and terminal use benchmark. Harbor/Terminus-2 harness; 3h timeout, 32 CPU/48 GB RAM; temp=1.0, top_p=0.95, top_k=20, max_...
QwenClawBench
Agentic Coding
QwenClawBench - internal real-user-distribution Claw agent benchmark (open-sourcing soon); temp=0.6, 256K ctx.
SWE Atlas - RF
Agentic Coding
SWE Atlas RF - agentic coding subtask; mini SWE agent harness, temp=1.0, top_p=0.95, 128K context, avg of 5 runs.
SWE Atlas - TW
Agentic Coding
SWE Atlas TW - agentic coding subtask; mini SWE agent harness, temp=1.0, top_p=0.95, 128K context, avg of 5 runs.
DeepSWE 1.1
Agentic Coding
DeepSWE 1.1 - agentic coding benchmark evaluated with Claude Code harness at temp=1.0, top_p=0.95, 256K context window.
QwenSWEBench
Agentic Coding
QwenSWEBench - in-house coding benchmark for evaluating software engineering capabilities. Evaluated with Claude Code harness, avg@3, 8-hour timeout, max_tok...
SWE-MM
Agentic Coding
SWE-MM - multimodal software engineering benchmark. Evaluated on Claude Code harness using public dev split of SWE-bench Multimodal with modifications.
NL2Repo-Bench
Agentic Coding
NL2Repo-Bench - repository generation from natural language. Evaluated with Claude Code harness. Bash commands that access the specific repository are disabl...
FrontierSWE
Agentic Coding
FrontierSWE - agentic coding benchmark. MEAN@5 scores from the official FrontierSWE leaderboard. Dominance scores recomputed from raw scores using the offici...
MLS-Bench-Lite
Agentic Coding
MLS-Bench-Lite - agentic coding benchmark. Evaluated with Claude Code using a 5-hour timeout and max_tokens=131,072. Scores from the official leaderboard.
AndroidBench
Agentic Coding
AndroidBench - evaluated on the 95-task public subset, reporting avg@3 scores.
QwenQoderBench
Agentic Coding
QwenQoderBench - in-house coding benchmark to evaluate user experience on Qoder. Evaluated with Claude Code harness, avg@5, 6-hour timeout, max_tokens=32768,...
QwenReactBench
Agentic Coding
QwenReactBench - in-house React project building benchmark using Claude Code harness. Bilingual (EN/CN), 7 categories; auto-render + multimodal judge; BT/Elo...
QwenSVGBench
Agentic Coding
QwenSVGBench - in-house SVG code generation benchmark; bilingual (EN/CN), auto-render + multimodal judge; BT/Elo rating.
PaperBench
Agentic Coding
PaperBench - evaluated in the BasicAgent setting under Code-Dev mode, judged by Claude Opus 4.6, averaged over 3 runs (max 12 hours per run).
WildClawBench
Agentic Coding
WildClawBench - agentic coding evaluation benchmark. Overall score from internlm/WildClawBench.
Terminal-Bench 2.1 (Best Reported Harness)
Agentic Coding
Terminal-Bench 2.1 (Best Reported Harness) - best score across all evaluation harnesses (Terminus-2, Claude Code, etc.) for Terminal-Bench 2.1.
FullStackBench en
Agentic Coding
FullStackBench en - English full-stack code generation benchmark.
FullStackBench zh
Agentic Coding
FullStackBench zh - Chinese full-stack code generation benchmark.
DeepSWE
Agentic Coding
DeepSWE - agentic coding benchmark from datacurve/deep-swe dataset evaluating software engineering agent capabilities.
DSBench-FullStack
Agentic Coding
DSBench-FullStack - internal full-stack development test set for evaluating coding agents on full-stack development tasks.
ProgramBench
Agentic Coding
ProgramBench - agentic coding benchmark evaluated with Claude-Code 2.1.156, 200 instances, temp=1.0, top_p=1.0, max_tokens=64000, max_turns=2000, sample_time...
Aider
Agentic Coding
Aider - coding benchmark based on the Aider coding assistant tool
Frontier-Bench v0.1
Agentic Coding
Frontier-Bench v0.1 - frontier coding benchmark evaluating advanced agentic coding capabilities.
Kimi Code Bench V2
Agentic Coding
In-house benchmark by Moonshot AI designed to evaluate coding agents on realistic tasks. It has diverse software engineering tasks across 10+ mainstream prog...
LHTB Solved
Agentic Coding
Long-Horizon-Terminal-Bench (LHTB) Solved metric from IntelligenceLab. Evaluates long-horizon terminal-based agentic tasks.
DSBench-Hard
Agentic Coding
DSBench-Hard - internal test set of difficult coding-agent problems for evaluating agentic coding on challenging software engineering tasks.
Program Bench
Agentic Coding
Evaluates code-generation agents by asking them to recreate a program's behavior from only a compiled binary and its documentation. It spans 200 tasks, from ...
Z.ai Code Bench
Agentic Coding
Z.AI's in-house coding benchmark for the GLM series. GLM-5.3 shows a 50% relative improvement over GLM-5.2; absolute scores are not published on the model ca...
Harbor-Index
Agentic Coding
Agentic coding index benchmark reported in the Tencent Hy4 preview model card benchmark appendix.
Hy-Backend 2.0 (Internal)
Agentic Coding
Tencent internal (Hy team) backend engineering benchmark, reported in the Hy4 preview model card appendix.
Hy-SWE Max Verified (Internal)
Agentic Coding
Tencent internal 300-task agentic coding benchmark; Claude Code scaffold, 200-turn budget, 16 CPU/32 GB sandbox, no web access, avg of 3 runs.
E-Bench-Code (Internal)
Agentic Coding
Tencent internal (Hy team) enterprise coding benchmark, reported in the Hy4 preview model card appendix.
Terminal-Bench 4.0
Agentic Coding
Terminal-Bench 4.0 agentic terminal tasks. DeepSeek-V4.1-Flash evaluated with the Minimal mode of DeepSeek Harness and a 1M-token context window; Pass@1.
CFEval
Agentic Coding
CFEval - Qwen's CodeForces-style competitive programming evaluation; score is an Elo-like rating (higher is better).
SWE-bench Verified (OpenCode)
Agentic Coding
SWE-bench Verified evaluated with the OpenCode agent harness. As evaluated on the LFM2.5-24B-A2B model card: vLLM, temperature = 0.6, max output tokens = 327...