Atlas

Benchmarks

← All benchmarks

Long-Horizon Terminal-Bench 1.0 (strict pass, reward = 1.0)

Code · 2026-07-09

The share of the 46 Long-Horizon Terminal-Bench 1.0 tasks on which an agent earns the full reward of 1.0 from the hidden verifier, in percent. It comes from the same single 90-minute Terminus 2 attempt per task as the mean-reward headline.

Benchmark source: arXiv

Top models (higher is better)

ModelScoreSource
Grok 4.519.6LHTB 1.0 leaderboard (Grok 4.5) · Tencent
Claude Fable 515.2LHTB 1.0 leaderboard (Claude Fable 5) · Tencent
Opus 4.810.9LHTB 1.0 leaderboard (Claude Opus 4.8) · Tencent
Sonnet 510.9LHTB 1.0 leaderboard (Claude Sonnet 5) · Tencent
Gemini 3.6 Flash10.9LHTB 1.0 leaderboard (Gemini 3.6 Flash) · Tencent
GPT-5.510.9LHTB 1.0 leaderboard (GPT-5.5) · Tencent
Kimi K310.9LHTB 1.0 leaderboard (Kimi K3) · Tencent
GPT-5.6 Sol8.7LHTB 1.0 leaderboard (GPT-5.6-sol) · Tencent
MiniMax M36.5LHTB 1.0 leaderboard (MiniMax M3) · Tencent
Sonnet 4.64.3LHTB 1.0 leaderboard (Claude Sonnet 4.6) · Tencent
Kimi K2.7 Code4.3LHTB 1.0 leaderboard (Kimi K2.7 Code) · Tencent
GPT-5.3-Codex2.2LHTB 1.0 leaderboard (GPT-5.3 Codex) · Tencent
GPT-5.42.2LHTB 1.0 leaderboard (GPT-5.4) · Tencent
DeepSeek-V4-Pro0.0LHTB 1.0 leaderboard (DeepSeek V4 Pro) · Tencent
Doubao Seed 2.1 Pro0.0LHTB 1.0 leaderboard (Doubao Seed 2.1 Pro) · Tencent
Gemini 3.1 Pro Preview0.0LHTB 1.0 leaderboard (Gemini 3.1 Pro) · Tencent
GLM-5.10.0LHTB 1.0 leaderboard (GLM 5.1) · Tencent
GLM-5.20.0LHTB 1.0 leaderboard (GLM 5.2) · Tencent
Grok 4.200.0LHTB 1.0 leaderboard (Grok 4.20) · Tencent
Kimi K2.60.0LHTB 1.0 leaderboard (Kimi K2.6) · Tencent
Qwen3.6 Plus (2026-04-02)0.0LHTB 1.0 leaderboard (Qwen3.6 Plus) · Tencent
Qwen3.7-Max (source unspecified)0.0LHTB 1.0 leaderboard (Qwen3.7 Max) · Tencent
Loading Atlas data…