Frontier Log

Benchmark · InactiveSuperseded by Agents' Last Exam v1.1

Agents' Last Exam v1

Created by UC Berkeley RDI, 3 Jun 2026 · Higher is better · Website

0204060Apr 2026Jul 2026Oct 202659.3%GPT-6 Astra26.3%GLM-5.3-Flash53.6%GPT-5.6 Sol20.4%GLM-5.246.9%GPT-5.50204060Apr 2026Jul 2026Oct 202659.3%GPT-6 Astra26.3%GLM-5.3-Flash53.6%GPT-5.6 Sol20.4%GLM-5.246.9%GPT-5.5

Results

ModelScoreSourceSource check
GPT-6 Astra3 Sep 2026 · OpenAI59.3%GPT-6 Astra: A new generation of intelligencebest of any effort; harness not stated
GLM-5.3-Flash26 Aug 2026 · Z.ai26.3%GLM-5.3-Flash: Frontier Intelligence, Flash CostClaude Code harness; max effort; 1M context; subset not statedOur capture
CorroboratedThe same number is printed in 3 other lab documents: 1 2 3
GPT-5.6 Luna9 Jul 2026 · OpenAI50.3%GPT-5.6: Frontier intelligence that scales with your ambitioneffort and harness not stated
GPT-5.6 Sol9 Jul 2026 · OpenAI53.6%GPT-5.6: Frontier intelligence that scales with your ambitionheadline figure; effort and harness not stated
CorroboratedThe same number is printed in 2 other lab documents: 1 2
GPT-5.6 Sol9 Jul 2026 · OpenAI52.7%GPT-5.6: Frontier intelligence that scales with your ambitioneval table figure; effort and harness not stated; another report, not charted
GPT-5.6 Terra9 Jul 2026 · OpenAI50.4%GPT-5.6: Frontier intelligence that scales with your ambitioneffort and harness not stated
GLM-5.216 Jun 2026 · Z.ai20.4%GLM-5.3-Flash: Frontier Intelligence, Flash CostClaude Code harness; subset not stated; reported laterOur capture
CorroboratedThe same number is printed in 3 other lab documents: 1 2 3
GPT-5.523 Apr 2026 · OpenAI46.9%GPT-5.6: Frontier intelligence that scales with your ambitioneffort and harness not stated; reported later

All benchmarks