Language modelActive

GPT-5.5

OpenAI

Released: April 2026
Data date: July 30, 2026

Category view

Position within the category

This overview uses only published data from matching cohorts. Missing values never change a rank.

The index averages rank percentiles from 4 documented comparison cohorts. Only models with complete coverage receive a position.

Position: Rank 13 of 26
Index score: 57.2 / 100
Coverage: 4 / 4

Leaderboard

1Claude Opus 588.9 / 100

2Gemini 3.8 Flash88.5 / 100

3Claude Fable 583.2 / 100

12Claude Sonnet 557.7 / 100

13GPT-5.557.2 / 100

14GLM-5.354.8 / 100

25MiMo-V2.5-Pro12.5 / 100

Source: Vals AI, GPQA Diamond Source: Vals AI, Vibe Code Bench v1.1 Source: Vals AI, Terminal-Bench 2.1 Source: Vals AI, Finance Agent (v2)

Measurements

Comparable benchmark results

Each chart contains exactly one source, one measurement series, and one stored comparison cohort. Bars show the position. The measured value appears on the right.

τ-bench Airline Clean

79.5% · Rank 3 of 3

Task: Accuracy · Comparison cohort: hal-tau-airline-clean-hal-reliability-overall · Data date: August 5, 2026

1.Gemini 3.1 Pro Preview82.1%

2.Gemini 3.5 Flash80.8%

3.GPT-5.579.5%

3 of 3 model versions shown in this chart. A higher value ranks first.

26 cleaned airline tasks. Princeton HAL recomputed every metric from scratch on this subset.

Source: Princeton HALParticipants: 3

τ-bench Airline Clean

0.89 rating · Rank 1 of 3

Task: Reliability · Comparison cohort: hal-tau-airline-clean-hal-reliability-reliability · Data date: August 5, 2026

1.GPT-5.50.89 rating

2.Gemini 3.1 Pro Preview0.86 rating

2.Gemini 3.5 Flash0.86 rating

3 of 3 model versions shown in this chart. A higher value ranks first.

26 cleaned airline tasks. Princeton HAL recomputed every metric from scratch on this subset.

Source: Princeton HALParticipants: 3

τ-bench Airline Clean

0.84 rating · Rank 1 of 3

Task: Consistency · Comparison cohort: hal-tau-airline-clean-hal-reliability-consistency · Data date: August 5, 2026

1.GPT-5.50.84 rating

2.Gemini 3.1 Pro Preview0.81 rating

3.Gemini 3.5 Flash0.78 rating

3 of 3 model versions shown in this chart. A higher value ranks first.

26 cleaned airline tasks. Princeton HAL recomputed every metric from scratch on this subset.

Source: Princeton HALParticipants: 3

τ-bench Airline Clean

0.84 rating · Rank 1 of 3

Task: Predictability · Comparison cohort: hal-tau-airline-clean-hal-reliability-predictability · Data date: August 5, 2026

1.GPT-5.50.84 rating

2.Gemini 3.1 Pro Preview0.83 rating

3.Gemini 3.5 Flash0.82 rating

3 of 3 model versions shown in this chart. A higher value ranks first.

26 cleaned airline tasks. Princeton HAL recomputed every metric from scratch on this subset.

Source: Princeton HALParticipants: 3

τ-bench Airline Clean

0.97 rating · Rank 1 of 3

Task: Robustness · Comparison cohort: hal-tau-airline-clean-hal-reliability-robustness · Data date: August 5, 2026

1.Gemini 3.5 Flash0.97 rating

1.GPT-5.50.97 rating

3.Gemini 3.1 Pro Preview0.94 rating

3 of 3 model versions shown in this chart. A higher value ranks first.

26 cleaned airline tasks. Princeton HAL recomputed every metric from scratch on this subset.

Source: Princeton HALParticipants: 3

τ-bench Airline Clean

0.96 rating · Rank 3 of 3

Task: Safety · Comparison cohort: hal-tau-airline-clean-hal-reliability-safety · Data date: August 5, 2026

1.Gemini 3.5 Flash0.99 rating

2.Gemini 3.1 Pro Preview0.97 rating

3.GPT-5.50.96 rating

3 of 3 model versions shown in this chart. A higher value ranks first.

26 cleaned airline tasks. Princeton HAL recomputed every metric from scratch on this subset.

Source: Princeton HALParticipants: 3

GAIA Reliability

62.8% · Rank 3 of 3

Task: Accuracy · Comparison cohort: hal-gaia-reliability-hal-reliability-overall · Data date: August 6, 2026

1.Gemini 3.5 Flash79.2%

2.Gemini 3.1 Pro Preview76.2%

3.GPT-5.562.8%

3 of 3 model versions shown in this chart. A higher value ranks first.

165 GAIA tasks with repeated runs and perturbations. HAL does not publish cost for this evaluation.

Source: Princeton HALParticipants: 3

GAIA Reliability

0.79 rating · Rank 3 of 3

Task: Reliability · Comparison cohort: hal-gaia-reliability-hal-reliability-reliability · Data date: August 6, 2026

1.Gemini 3.5 Flash0.84 rating

2.Gemini 3.1 Pro Preview0.82 rating

3.GPT-5.50.79 rating

3 of 3 model versions shown in this chart. A higher value ranks first.

165 GAIA tasks with repeated runs and perturbations. HAL does not publish cost for this evaluation.

Source: Princeton HALParticipants: 3

GAIA Reliability

0.61 rating · Rank 3 of 3

Task: Consistency · Comparison cohort: hal-gaia-reliability-hal-reliability-consistency · Data date: August 6, 2026

1.Gemini 3.1 Pro Preview0.76 rating

2.Gemini 3.5 Flash0.75 rating

3.GPT-5.50.61 rating

3 of 3 model versions shown in this chart. A higher value ranks first.

165 GAIA tasks with repeated runs and perturbations. HAL does not publish cost for this evaluation.

Source: Princeton HALParticipants: 3

GAIA Reliability

0.8 rating · Rank 1 of 3

Task: Predictability · Comparison cohort: hal-gaia-reliability-hal-reliability-predictability · Data date: August 6, 2026

1.Gemini 3.5 Flash0.8 rating

1.GPT-5.50.8 rating

3.Gemini 3.1 Pro Preview0.78 rating

3 of 3 model versions shown in this chart. A higher value ranks first.

165 GAIA tasks with repeated runs and perturbations. HAL does not publish cost for this evaluation.

Source: Princeton HALParticipants: 3

GAIA Reliability

0.95 rating · Rank 2 of 3

Task: Robustness · Comparison cohort: hal-gaia-reliability-hal-reliability-robustness · Data date: August 6, 2026

1.Gemini 3.5 Flash0.96 rating

2.GPT-5.50.95 rating

3.Gemini 3.1 Pro Preview0.94 rating

3 of 3 model versions shown in this chart. A higher value ranks first.

165 GAIA tasks with repeated runs and perturbations. HAL does not publish cost for this evaluation.

Source: Princeton HALParticipants: 3

GAIA Reliability

1 rating · Rank 1 of 3

Task: Safety · Comparison cohort: hal-gaia-reliability-hal-reliability-safety · Data date: August 6, 2026

1.Gemini 3.1 Pro Preview1 rating

1.Gemini 3.5 Flash1 rating

1.GPT-5.51 rating

3 of 3 model versions shown in this chart. A higher value ranks first.

165 GAIA tasks with repeated runs and perturbations. HAL does not publish cost for this evaluation.

Source: Princeton HALParticipants: 3

AlmanBench v0.1

95.43% · Rank 1 of 5

Task: Accepted cases · Comparison cohort: almanbench-v0-1:official:overall · Data date: August 9, 2026

1.GPT-5.595.43%

2.GPT-5.6 Sol95.14%

3.Claude Opus 594.95%

4.DeepSeek-V4-Flash93%

5.Kimi K390.96%

5 of 5 model versions shown in this chart. A higher value ranks first.

1,029 public tasks with one direct run per row. The target is the Alman language specification.

Source: Alman InstitutParticipants: 5

SWE-Marathon v1.0

12% · Rank 3 of 4

Task: Binary resolution, pass@1 · Comparison cohort: swe-marathon-v1-0:official:pass1 · Data date: August 9, 2026

1.Grok 4.529%

2.Claude Fable 524%

3.GPT-5.512%

4.Gemini 3.5 Flash7%

4 of 4 model versions shown in this chart. A higher value ranks first.

20 multi-hour tasks. The named coding agent is part of every result row.

Source: SWE-MarathonParticipants: 4

LiveBench

80.19 points · Rank 3 of 4

Comparison cohort: livebench-2026-06-25-max · Data date: September 2, 2026

1.Claude Fable 582.97 points

2.GPT-5.6 Sol81.05 points

3.GPT-5.580.19 points

4.GPT-5.6 Terra77.94 points

4 of 4 model versions shown in this chart. A higher value ranks first.

23 objectively scored tasks across seven categories.

Source: LiveBenchParticipants: 4

LiveBench

89.7 points · Rank 3 of 4

Task: Reasoning · Comparison cohort: livebench-2026-06-25-reasoning-max · Data date: September 2, 2026

1.GPT-5.6 Sol91.7 points

2.GPT-5.6 Terra90.6 points

3.Claude Fable 589.7 points

3.GPT-5.589.7 points

4 of 4 model versions shown in this chart. A higher value ranks first.

Reasoning category inside the same LiveBench release.

Source: LiveBenchParticipants: 4

LiveBench

82.1 points · Rank 3 of 4

Task: Coding · Comparison cohort: livebench-2026-06-25-coding-max · Data date: September 2, 2026

1.Claude Fable 586 points

2.GPT-5.6 Sol83.9 points

3.GPT-5.582.1 points

4.GPT-5.6 Terra78.2 points

4 of 4 model versions shown in this chart. A higher value ranks first.

Coding category inside the same LiveBench release.

Source: LiveBenchParticipants: 4

LiveBench

54 points · Rank 4 of 4

Task: Agentic coding · Comparison cohort: livebench-2026-06-25-agentic-coding-max · Data date: September 2, 2026

1.Claude Fable 562.2 points

2.GPT-5.6 Sol56.2 points

3.GPT-5.6 Terra54.9 points

4.GPT-5.554 points

4 of 4 model versions shown in this chart. A higher value ranks first.

Agentic coding category inside the same LiveBench release.

Source: LiveBenchParticipants: 4

LiveBench

95.9 points · Rank 3 of 4

Task: Mathematics · Comparison cohort: livebench-2026-06-25-mathematics-max · Data date: September 2, 2026

1.GPT-5.6 Sol96.2 points

2.Claude Fable 596 points

3.GPT-5.595.9 points

4.GPT-5.6 Terra94.9 points

4 of 4 model versions shown in this chart. A higher value ranks first.

Mathematics category inside the same LiveBench release.

Source: LiveBenchParticipants: 4

LiveBench

81.6 points · Rank 1 of 4

Task: Data analysis · Comparison cohort: livebench-2026-06-25-data-analysis-max · Data date: September 2, 2026

1.GPT-5.581.6 points

2.Claude Fable 580.5 points

3.GPT-5.6 Sol79.8 points

4.GPT-5.6 Terra79.3 points

4 of 4 model versions shown in this chart. A higher value ranks first.

Data analysis category inside the same LiveBench release.

Source: LiveBenchParticipants: 4

LiveBench

87.4 points · Rank 3 of 4

Task: Language · Comparison cohort: livebench-2026-06-25-language-max · Data date: September 2, 2026

1.Claude Fable 590.7 points

2.GPT-5.6 Sol87.7 points

3.GPT-5.587.4 points

4.GPT-5.6 Terra82.9 points

4 of 4 model versions shown in this chart. A higher value ranks first.

Language category inside the same LiveBench release.

Source: LiveBenchParticipants: 4

LiveBench

70.7 points · Rank 3 of 4

Task: Instruction following · Comparison cohort: livebench-2026-06-25-instruction-following-max · Data date: September 2, 2026

1.Claude Fable 575.8 points

2.GPT-5.6 Sol71.8 points

3.GPT-5.570.7 points

4.GPT-5.6 Terra64.6 points

4 of 4 model versions shown in this chart. A higher value ranks first.

Instruction following category inside the same LiveBench release.

Source: LiveBenchParticipants: 4

BrowseComp

84.4% · Rank 4 of 5

Comparison cohort: browsecomp-openai-gpt-5-6-table · Data date: September 2, 2026

1.Claude Mythos 588%

2.GPT-5.6 Terra87.5%

3.Gemini 3.1 Pro Preview85.9%

4.GPT-5.584.4%

5.GPT-5.6 Luna83.3%

5 of 5 model versions shown in this chart. A higher value ranks first.

Cross-vendor comparison table with browsing tools. Exact agent systems differ.

Editorial selection from the vendor table, not a complete extract of the comparison cohort.

Source: OpenAIParticipants: 5

Individual values

Published individual values

No exactly matching published comparison cohort is available for these values. The bar shows only the documented scale, not a rank.

OSWorld 2.0

Data date: August 26, 2026

OSWorld 2.013%

Scale 0 to 100. Not ranked

Complete agent system with up to 500 steps.

Claude used about 244,000 output tokens. This is a system result, not an isolated model score.

Source: OSWorld 2.0

SWE-Bench Pro

Data date: September 2, 2026

SWE-Bench Pro58.6%

Scale 0 to 100. Not ranked

Comparison table published by Anthropic. Not a Gradually test.

After an audit, OpenAI estimates that about 30% of the public tasks are broken. The result therefore remains a disputed secondary signal. The rows are an editorial selection from the respective comparison table.

Source: Anthropic

Profile

Specifications and access

Published information about this model. Unknown values are not estimated.

Model type: ProprietarySource
Context window: 1,050,000 tokensSource
Knowledge cutoff: December 1, 2025Source
Notes: 1M context, agentic workflows, Terminal-Bench 2.0 82.7%, parameter count not disclosedSource