Frontier Log

Tencent

Hy4

Released · tencent.ai

Price

Per 1M tokensInputCachedOutput
List price, billed at 50% until 31 Dec 2026 (Beijing time)$0.834$0.042$2.501
Price history
DateInputCachedOutput
hy.tencent.ai · 2 · Checked$0.834$0.042$2.501

Context window

TokensContext windowInput limitOutput limit
Default1M960K64K
Context history
DateContext windowInput limitOutput limit
hy.tencent.ai · 2 · Corroborated1M960K64K

Benchmark results

BenchmarkScoreSource verified
Humanity's Last ExamWith tools · full set; highest reasoning setting59.1%Checked A person checked this number against its source on 5 Nov 2026. This confirms the number matches what the lab published, not that the score was reproduced.
Hy4 preview, re-evaluated at Hy4 launch; text-only subset57.3%Checked A person checked this number against its source on 5 Nov 2026. This confirms the number matches what the lab published, not that the score was reproduced.
Terminal-Bench 4.0Claude Code harness; up to 500 turns and a 13h timeout per trial; highest reasoning setting44.4%Checked A person checked this number against its source on 5 Nov 2026. This confirms the number matches what the lab published, not that the score was reproduced.
Hy4 preview, re-evaluated at Hy4 launch; Claude Code harness; up to 500 turns and a 13h timeout per trial26.3%Checked A person checked this number against its source on 5 Nov 2026. This confirms the number matches what the lab published, not that the score was reproduced.
Terminal-Bench Science 0.1Claude Code harness; up to 500 turns and a 13h timeout per trial; highest reasoning setting19.8%Checked A person checked this number against its source on 5 Nov 2026. This confirms the number matches what the lab published, not that the score was reproduced.
Hy4 preview, re-evaluated at Hy4 launch; Claude Code harness; up to 500 turns and a 13h timeout per trial16.2%Checked A person checked this number against its source on 5 Nov 2026. This confirms the number matches what the lab published, not that the score was reproduced.
AutomationBenchAutomationBench v1.0.6; 600 public tasks; official method; highest reasoning setting56.2%Checked A person checked this number against its source on 5 Nov 2026. This confirms the number matches what the lab published, not that the score was reproduced.
Hy4 preview, re-evaluated at Hy4 launch; AutomationBench v1.0.6; 600 public tasks; official method36.3%Checked A person checked this number against its source on 5 Nov 2026. This confirms the number matches what the lab published, not that the score was reproduced.
OSWorld 2.1Full set, partial score · harness not stated; highest reasoning setting78.9%Checked A person checked this number against its source on 5 Nov 2026. This confirms the number matches what the lab published, not that the score was reproduced.
DeepSWE v1.1mini-swe-agent harness; highest reasoning setting74.9%Checked A person checked this number against its source on 5 Nov 2026. This confirms the number matches what the lab published, not that the score was reproduced.
Agents' Last Exam v1.1Pass rate · harness and subset not stated; highest reasoning setting41.0%Checked A person checked this number against its source on 5 Nov 2026. This confirms the number matches what the lab published, not that the score was reproduced.
Sources

Reported by Tencent

Related

All models