Skip to content
aiqbench

Capability frontier over time

Each Epoch benchmark, plotted on the release date of every model that's been scored on it. As newer models arrive the ceiling of what's been achieved keeps rising — a real trend, distinct from and often mistaken for individual models improving over time.

BenchmarkGPQA DiamondOTIS Mock AIMEFrontierMathSimpleQA VerifiedSWE-bench VerifiedARC-AGI-2Terminal-BenchHumanity's Last ExamSimpleBench

FrontierMath — models on arrival

One point per model at its release date. The line traces the best-known score at each moment.

0%25%50%75%100%2024202520262027Model release dateScore (%)gpt-4o-2024-08-06 Released: May 2024 Score: 0.3% 95% CI: [0.0%, 0.7%] Date source: curatedclaude-3-5-sonnet-20241022 Released: Jun 2024 Score: 2.1% 95% CI: [1.2%, 2.9%] Date source: curatedo1-mini-2024-09-12 Released: Sep 2024 Score: 1.7% 95% CI: [1.0%, 2.5%] Date source: curatedgemini-1.5-flash-002 Released: Sep 2024 Score: 0.0% 95% CI: [0.0%, 0.0%] Date source: curatedmistral-large-2411 Released: Nov 2024 Score: 0.3% 95% CI: [0.0%, 0.7%] Date source: name-suffixclaude-3-5-haiku-20241022 Released: Nov 2024 Score: 0.3% 95% CI: [0.0%, 0.7%] Date source: curatedo1-2024-12-17 Released: Dec 2024 Score: 9.3% 95% CI: [7.6%, 11.0%] Date source: curatedDeepSeek-V3 Released: Dec 2024 Score: 1.7% 95% CI: [1.0%, 2.5%] Date source: curatedo3-mini-2025-01-31 Released: Jan 2025 Score: 12.4% 95% CI: [10.5%, 14.4%] Date source: curatedo3-minigemini-2.0-flash-001 Released: Feb 2025 Score: 1.7% 95% CI: [1.0%, 2.5%] Date source: curatedgrok-3-mini-beta Released: Feb 2025 Score: 5.9% 95% CI: [4.5%, 7.2%] Date source: curatedclaude-3-7-sonnet-20250219 Released: Feb 2025 Score: 4.1% 95% CI: [3.0%, 5.3%] Date source: curatedLlama-4-Scout-17B-16E-Instruct Released: Apr 2025 Score: 0.0% 95% CI: [0.0%, 0.0%] Date source: curatedgpt-4.1-mini-2025-04-14 Released: Apr 2025 Score: 4.5% 95% CI: [3.3%, 5.7%] Date source: curatedgpt-4.1-2025-04-14 Released: Apr 2025 Score: 5.5% 95% CI: [4.2%, 6.9%] Date source: curatedgpt-4.1-nano-2025-04-14 Released: Apr 2025 Score: 1.0% 95% CI: [0.4%, 1.6%] Date source: curatedo4-mini-2025-04-16 Released: Apr 2025 Score: 24.8% 95% CI: [22.3%, 27.4%] Date source: curatedo4-minio3-2025-04-16 Released: Apr 2025 Score: 18.7% 95% CI: [16.4%, 21.0%] Date source: curatedqwen-plus-2025-04-28 Released: Apr 2025 Score: 1.7% 95% CI: [1.0%, 2.5%] Date source: name-suffixmistral-medium-2505 Released: May 2025 Score: 0.3% 95% CI: [0.0%, 0.7%] Date source: name-suffixclaude-opus-4-20250514 Released: May 2025 Score: 4.5% 95% CI: [3.3%, 5.7%] Date source: curatedclaude-sonnet-4-20250514 Released: May 2025 Score: 4.1% 95% CI: [3.0%, 5.3%] Date source: curatedgemini-2.5-flash Released: Jun 2025 Score: 4.8% 95% CI: [3.6%, 6.1%] Date source: curatedgemini-2.5-pro Released: Jun 2025 Score: 14.1% 95% CI: [12.1%, 16.2%] Date source: curatedgemini-2.5-pro-preview-06-05 Released: Jun 2025 Score: 10.3% 95% CI: [8.6%, 12.1%] Date source: curatedgrok-4-0709 Released: Jul 2025 Score: 19.7% 95% CI: [17.3%, 22.0%] Date source: curatedclaude-opus-4-1-20250805 Released: Aug 2025 Score: 7.2% 95% CI: [5.7%, 8.8%] Date source: curatedgpt-5-2025-08-07 Released: Aug 2025 Score: 32.4% 95% CI: [29.7%, 35.2%] Date source: curatedgpt-5gpt-5-mini-2025-08-07 Released: Aug 2025 Score: 27.2% 95% CI: [24.6%, 29.9%] Date source: curatedgpt-5-nano-2025-08-07 Released: Aug 2025 Score: 8.3% 95% CI: [6.7%, 9.9%] Date source: curatedclaude-sonnet-4-5-20250929 Released: Sep 2025 Score: 15.2% 95% CI: [13.1%, 17.3%] Date source: curatedclaude-haiku-4-5-20251001 Released: Oct 2025 Score: 5.9% 95% CI: [4.5%, 7.3%] Date source: curatedclaude-opus-4-5-20251101 Released: Nov 2025 Score: 20.7% 95% CI: [18.3%, 23.1%] Date source: name-suffixgpt-5.1-2025-11-13 Released: Nov 2025 Score: 31.0% 95% CI: [28.3%, 33.8%] Date source: name-suffixgpt-5.2-2025-12-11 Released: Dec 2025 Score: 40.7% 95% CI: [37.8%, 43.6%] Date source: name-suffixgpt-5.2glm-5 Released: Feb 2026 Score: 16.4% 95% CI: [14.2%, 18.6%] Date source: huggingface-createdqwen3.5-plus Released: Feb 2026 Score: 21.0% 95% CI: [18.6%, 23.4%] Date source: name-suffixqwen3.5-flash Released: Feb 2026 Score: 6.2% 95% CI: [4.8%, 7.6%] Date source: name-suffixgpt-5.4-pro-2026-03-05 Released: Mar 2026 Score: 50.0% 95% CI: [47.1%, 52.9%] Date source: name-suffixgpt-5.4-proglm-5.1 Released: Apr 2026 Score: 33.4% 95% CI: [30.7%, 36.2%] Date source: huggingface-createdkimi-k2.6 Released: Apr 2026 Score: 39.0% 95% CI: [36.1%, 41.8%] Date source: huggingface-createdclaude-opus-4-6 Released: Sep 2026 Score: 40.7% 95% CI: [37.8%, 43.6%] Date source: lmarena-first-seengemini-3.5-flash Released: Sep 2026 Score: 39.0% 95% CI: [36.1%, 41.8%] Date source: lmarena-first-seengpt-5.4-2026-03-05 Released: Sep 2026 Score: 47.6% 95% CI: [44.7%, 50.5%] Date source: lmarena-first-seenclaude-opus-4-7 Released: Sep 2026 Score: 43.8% 95% CI: [40.9%, 46.7%] Date source: lmarena-first-seenmuse-spark Released: Sep 2026 Score: 39.0% 95% CI: [36.1%, 41.9%] Date source: lmarena-first-seenclaude-opus-4-8 Released: Sep 2026 Score: 47.2% 95% CI: [44.3%, 50.2%] Date source: lmarena-first-seenclaude-sonnet-4-6 Released: Sep 2026 Score: 32.4% 95% CI: [29.6%, 35.2%] Date source: lmarena-first-seengemini-3.1-pro-preview Released: Sep 2026 Score: 36.9% 95% CI: [34.1%, 39.7%] Date source: lmarena-first-seengpt-5.4-mini-2026-03-17 Released: Oct 2026 Score: 28.3% 95% CI: [26.2%, 30.3%] Date source: lmarena-first-seenqwen3.6-plus Released: Oct 2026 Score: 26.2% 95% CI: [23.6%, 28.8%] Date source: lmarena-first-seengpt-5.4-nano-2026-03-17 Released: Oct 2026 Score: 25.9% 95% CI: [23.3%, 28.4%] Date source: lmarena-first-seenqwen3.6-max-preview Released: Oct 2026 Score: 23.1% 95% CI: [20.6%, 25.6%] Date source: lmarena-first-seen
Model at its release dateRunning best score (frontier)Score confidence interval

Read this as a frontier chart, not a per-model trajectory. A fixed model's score on a fixed benchmark does not change over time — Claude 3.5 Sonnet's GPQA number today is the same one Anthropic reported at launch. The line rises because better models keep arriving, not because any specific model got better. Seeing one model's score change over time (as its evaluation is re-run with different scaffolds, for example) needs a different kind of data: daily snapshots of each board, which we will chart once several weeks of them have accumulated.

17 models have no known release date and are not plotted.

Benchmarks · frontier

Frontier milestones

Every time a ceiling moved: who, which benchmark, by how much. Tap a sparkline to pick a benchmark — the chart above follows it; tap an organisation to filter by lab. Each event is the diff of the running best the chart draws.

8 of 114 frontier events ·

  1. Mar 202650.0%

    gpt-5.4-pro-2026-03-05 OpenAI sets FrontierMath

    +9.3 pp
  2. Dec 202540.7%

    gpt-5.2-2025-12-11 OpenAI sets FrontierMath

    +8.3 pp
  3. Aug 202532.4%

    gpt-5-2025-08-07 OpenAI sets FrontierMath

    +7.6 pp
  4. Apr 202524.8%

    o4-mini-2025-04-16 OpenAI sets FrontierMath

    +12.4 pp
  5. Jan 202512.4%

    o3-mini-2025-01-31 OpenAI sets FrontierMath

    +3.1 pp
  6. Dec 20249.3%

    o1-2024-12-17 OpenAI sets FrontierMath

    +7.2 pp
  7. Jun 20242.1%

    claude-3-5-sonnet-20241022 Anthropic sets FrontierMath

    +1.7 pp
  8. May 20240.3%

    gpt-4o-2024-08-06 OpenAI sets FrontierMath

    first score