Kimi K2 Thinking

Terminal-Bench Hard (AA) pinned revision 74221fb → · Published board — Tests a pinned 44-task hard subset of terminal-based work using Terminus 2.

τ²-Bench Airline (OpenRouter run) published 2026-09-15 → · Published board — Multi-turn service agents making tool calls under strict policy constraints. (OpenRouter's own reproducible run, as published on openrouter.ai/benchmarks.)

SWE-bench Verified published 2026-09-10 → · mini-SWE-agent — A 500-instance human-filtered subset of SWE-bench created with OpenAI, served as the default Verified leaderboard view.

GPQA Diamond (OpenRouter run) — measured cost per task published 2026-09-15 → · Published board — Mean USD OpenRouter actually spent per task while running GPQA Diamond for this configuration.

τ²-Bench Airline (OpenRouter run) — measured cost per task published 2026-09-15 → · Published board — Mean USD OpenRouter actually spent per task while running τ²-Bench Airline for this configuration.

IFBench (AA single-turn) published 2026-09-10 → · Published board — Tests precise single-turn instructions with deterministic rule checks.

Humanity's Last Exam (AA text-only) published 2026-09-10 → · Published board — Tests expert-level knowledge on AA's text-only Humanity's Last Exam subset.

AA-Omniscience Index published 2026-09-10 → · Published board — Tests factual reliability while rewarding correct answers and penalizing hallucinations.

UGI NatInt published 2026-09-10 → · Published board — A private-question evaluation of general knowledge and reasoning across textbook, popular-culture and world-model topics.

AA-LCR v1.1 v1.1 → · Published board — Tests reasoning across multiple long documents with corrected answer keys and grading.

AIME 2025 (AA) v2025 → · Published board — Advanced mathematical problem solving on AIME I and II 2025.

HMMT November 2025 (MathArena) 2025-11 → · Published board — The 30 final-answer problems of the November 2025 Harvard-MIT Mathematics Tournament, one of the largest high-school math competitions in the United States, run by MathArena on the contest problems.

Apex 2025 (MathArena) v2025 → · Published board — Twelve final-answer problems from 2025 competitions that MathArena selected because strong models of the time failed them; the selection is biased against the four models used to choose it (Grok 4, GPT-5, Gemini 2.5 Pro, GLM 4.5).

Apex Shortlist 2025 (MathArena) v2025 → · Published board — 47 final-answer problems from 2025 competitions on which at least one of two reference models (Grok 4 Fast, GPT-5 mini) got at least one of four attempts wrong, run by MathArena.

SimpleBench published 2026-09-10 → · Published board — A multiple-choice text benchmark of over 200 questions covering spatio-temporal reasoning, social intelligence, and linguistic adversarial robustness, on which a non-specialized human baseline outperforms every tested LLM.

SimpleBench published 2026-09-26 → · Published board — A multiple-choice text benchmark of over 200 questions covering spatio-temporal reasoning, social intelligence, and linguistic adversarial robustness, on which a non-specialized human baseline outperforms every tested LLM.

CritPt (AA) published 2026-09-10 → · Published board — Tests research-level physics reasoning with Python, symbolic and numerical answers.

GPQA Diamond (AA) published 2026-09-10 → · Published board — Tests graduate-level biology, physics and chemistry knowledge on the Diamond subset.

GPQA Diamond (OpenRouter run) published 2026-09-15 → · Published board — Graduate-level science questions that resist retrieval and reward careful reasoning. (OpenRouter's own reproducible run, as published on openrouter.ai/benchmarks.)

τ²-Bench Telecom (AA) published 2026-09-10 → · Published board — Tests dual-control telecom agents that coordinate tool use with a simulated user.

Toolathlon, original release (HKUST NLP, archived board) pre-verified → · Published board — An agent completes 108 real-world tool-use tasks across MCP servers and local toolkits, graded by executing the resulting state against each task's ground truth — the original task definitions, before the June 2026 Verified repair.

UGI Willingness published 2026-09-10 → · Published board — A private-question evaluation of how far a model follows challenging instructions before refusing or deviating.

UGI published 2026-09-10 → · Published board — A private-question suite combining sensitive-topic knowledge with willingness to follow controversial instructions.

EQ-Bench Creative Writing v3 v3 → · Published board — A LLM-judged creative writing benchmark with vocabulary and GPT-slop controls.

EQ-Bench Longform Creative Writing v1.11 → · Published board — LLM-judged benchmark in which models plan and write a short story/novella over 8x 1000-word turns, scored 0-100 across 14 rubric dimensions.

UGI Writing published 2026-09-10 → · Published board — A maintainer-owned writing evaluation considering intelligence, style, repetition and length adherence, informed by human preferences.