Atlas

Benchmarks

TCSExam

GitHub

TCSExam (Tsinghua University, Moonshot AI and Shanghai Qizhi Institute) gives coding agents theoretical computer science research tasks with up to 12 hours each, checked by the Lean kernel or an executable checker: 20 Lean theorems, 30 solver-engineering problems scored against three fixed Kimi K3 runs, 20 constructions scored against a reference, and 30 open conjectures to prove or refute. The Index is 30P + 30S + 30C + 10O over the track means, in points; 100 matches the references everywhere, and constructions that beat their reference can push it past 100.

Models
8
Top score
71.9
Leader
GPT-6 Astra
Updated
Nov 5, 2026
Resolution
~1 mo
Sharp
Saturation
~Nov 2027
Projected
Signal
92ndpercentile
205 benchmarks from 0.8 to 3.6219 benchmarks from 3.6 to 6.4134 benchmarks from 6.4 to 9.393 benchmarks from 9.3 to 12.155 benchmarks from 12.1 to 14.925 benchmarks from 14.9 to 17.713 benchmarks from 17.7 to 20.55 benchmarks from 20.5 to 23.37 benchmarks from 23.3 to 26.27 benchmarks from 26.2 to 29.01 benchmark from 29.0 to 31.83 benchmarks from 31.8 to 34.66 benchmarks from 34.6 to 37.43 benchmarks from 37.4 to 40.21 benchmark from 40.2 to 43.11 benchmark from 45.9 to 48.71 benchmark from 65.6 to 68.4

61st strongest of 779

Signal-to-noise +16.16

Difficulty
72ndpercentile
1 benchmark from 5.9 to 14.81 benchmark from 23.6 to 32.52 benchmarks from 32.5 to 41.31 benchmark from 41.3 to 50.22 benchmarks from 50.2 to 59.08 benchmarks from 59.0 to 67.97 benchmarks from 67.9 to 76.819 benchmarks from 76.8 to 85.623 benchmarks from 85.6 to 94.541 benchmarks from 94.5 to 103.331 benchmarks from 103.3 to 112.239 benchmarks from 112.2 to 121.148 benchmarks from 121.1 to 129.973 benchmarks from 129.9 to 138.859 benchmarks from 138.8 to 147.6118 benchmarks from 147.6 to 156.5127 benchmarks from 156.5 to 165.3113 benchmarks from 165.3 to 174.235 benchmarks from 174.2 to 183.117 benchmarks from 183.1 to 191.99 benchmarks from 191.9 to 200.81 benchmark from 200.8 to 209.64 benchmarks from 209.6 to 218.5

215th hardest of 779

Capability needed 162.7

8 dated models · 2 frontier modelsClosed weightsOpen weights
Data: Moonshot AIatlas.village.dev
#ModelSettingScoreSource
1GPT-6 Astramax · Codex
71.9
TCSExam paper Table 2 (Full, GPT-6 Astra)
2Kimi K3.1 Promax · Kimi Code
57.4
TCSExam paper Table 2 (Full, Kimi K3.1 Pro)
3Kimi K3max · Kimi Code
46.8
TCSExam paper Table 2 (Full, Kimi K3)
4DeepSeek V4.1 Flash 1017max · Claude Code
46.0
TCSExam paper Table 2 (Full, DeepSeek V4.1 Flash 1017)
5GLM-5.3max · Claude Code
45.5
TCSExam paper Table 2 (Full, GLM-5.3)
6Qwen3.8 Max (0902)max · Claude Code
44.8
TCSExam paper Table 2 (Full, Qwen3.8 Max (0902))
7Grok 4.6x-high · Grok Build
40.5
TCSExam paper Table 2 (Full, Grok 4.6)
8DeepSeek V4.1 Flashmax · Claude Code
40.4
TCSExam paper Table 2 (Full, DeepSeek V4.1 Flash 0910)
Loading Atlas data…