Skip to main content
Radar

Five or fewer confirmed AI changes, with original sources, on mornings when something changed.A free source-linked morning brief.

See the free Radar Brief
Superseded.OpenAI has newer models in this line:GPT-5.6 Sol

GPT-5.5

SupersededReleased Apr 23, 2026ProprietaryReasoning1M context

Released Apr 23, 2026 see all recent releases

Decision reading
GPT-5.5 scores 73.3 out of 100 and ranks #8 of 232. This profile shows 42 source-displayable benchmark rows; its strongest eligible category is Knowledge at #7. API pricing is $5 input and $30 output per million tokens, with cached input at $0.5.

Data as of September 4, 2026 · How the score is built

Strongest published evidence

Knowledge ranks #7. Particularly effective for knowledge-intensive tasks like research, analysis, and factual Q&A.

Validate before choosing

42 published rows leave some tracked benchmark slots empty. Multimodal & Grounded is its lowest eligible category at #19.

Decision snapshot

Each value carries a field reference instead of floating alone. Markers compare this model with the current ranked and priced catalog; they are not absolute quality thresholds.

Capability

73.3/100

field median 58.1

#8 of 232 ranked models

Price

$5input / $30 output

input median $1

cached $0.50 · blended $17.50

Speed

84tok/s

field median 92 tok/s

First token 64.16 s

Context

1Mtokens

field median 256,000

Maximum output length is tracked separately

Capability shape

Each axis shows percentile within that category’s eligible cohort. The comparison outline is the median of the six nearest public-score peers; a collapsed vertex means the category is not rank-eligible.

GPT-5.5 category percentile values

  • Agentic91st percentile
  • Coding96th percentile
  • Reasoning33rd percentile
  • Knowledge97th percentile
  • MathNot eligible
  • MultilingualNot eligible
  • Multimodal62nd percentile
  • Instruction followingNot eligible

The dashed outline is median of 6 nearest peers.

Eligible category ranks

  1. Agentic#15/151
  2. Coding#8/183
  3. Reasoning#15/22
  4. Knowledge#7/181
  5. MathNot ranked
  6. MultilingualNot ranked
  7. Multimodal#19/48
  8. Inst. FollowingNot ranked
Top decileTop quartileMid-fieldNot eligible

What it costs to get this score

Published API price against the public score. The x-axis uses a log scale; the dashed path marks models that are not beaten by a cheaper, higher-scoring option. Price uses average of published input and output rates.

Explore all models

The chart opens on the current model. Scroll horizontally to inspect the full price axis.

Current modelGPT-5.5 · 73.3 score · $17.50 blended per million tokens
30405060708090$0.50$1$5$10$25↘ frontierGPT-5.5

Horizontal: blended price per million tokens, log scale · Vertical: public score

How much of this is verified

Coverage is split by category so a strong number never hides a thin evidence base. Verified means the row is tied to a published source; provisional rows remain visible but separate.

  1. Agentic13/13 verified
  2. Coding9/9 verified
  3. Reasoning4/4 verified
  4. Knowledge6/6 verified
  5. Math3/3 verified
  6. MultilingualNot measured
  7. Multimodal3/3 verified
  8. Inst. FollowingNot measured
Verified sourceProvisionalNot measured

Spec sheet

Each documented value carries its source. Missing fields stay visible as not sourced or not published, rather than disappearing from the page.

API model ID
gpt-5.5OpenAI pricing
Context window
1MOpenAI pricing
Maximum output
Not sourced yet
Knowledge cutoff
Not sourced yet
Input modalities
Not sourced yet
Output modalities
Not sourced yet
Parameters
Not disclosed by the provider
Availability
Not sourced yet
Cloud regions
Not tracked yet
Lifecycle
Superseded
API capabilities
Tool calling, structured outputs, and batch support are not tracked yet
Prompt caching
Published at $0.50 per million cached input tokensOpenAI pricing
Self-host
Weights are not published
Rate limits
Not tracked yet

Category score record

Scores and ranks appear only where published evidence can be displayed. The table keeps the score, weight, cohort, and evidence state together.

Category scores, ranks, weighting, benchmark coverage, and evidence status
CategoryScore
AgenticRank #15 of 151Percentile 91stWeight 22%13 benchmarksVerified63.9
CodingRank #8 of 183Percentile 96thWeight 20%9 benchmarksVerified67.7
ReasoningRank #15 of 22Percentile 33rdWeight 17%4 benchmarksVerified63.5
KnowledgeRank #7 of 181Percentile 97thWeight 12%6 benchmarksVerified73.3
MathRank Not rankedWeight 5%3 benchmarksVerified69.6
MultilingualWeight 7%0 benchmarksNot measuredNot measured
MultimodalRank #19 of 48Percentile 62ndWeight 12%3 benchmarksVerified71.3
Inst. FollowingWeight 5%0 benchmarksNot measuredNot measured

Benchmark ledger

Coding opens by default. The marker compares each value with the best source-verified result in the catalog; provisional leaders do not set the reference. Expand the remaining categories for every published row.

Coding9 rows
Coding benchmark values, best verified comparison, weight, and source status
SWE-bench ProScore58.6%Versus best verified row

Best verified: Claude Fable 5.1 · 81.2%

Gap22.6 behindWeightWeighted 25%
LiveCodeBench (Vals)LiveCodeBench, Vals AI runScore85.3%Versus best verified row

Best verified: Claude Fable 5.1 · 90.5%

Gap5.2 behindWeightWeighted 15%
cursorBench32Score58.4%Versus best verified row

Best verified: Claude Fable 5.1 · 73.4%

Gap15 behindWeightWeighted 10%
Terminal-Bench 2.0Score82.0%Versus best verified row

Best verified: GPT-5.6 Sol · 91.9%

Gap9.9 behindWeightDisplay only
Vibe Code BenchVibe Code Bench v1.1Score69.85%Versus best verified row

Best verified: Claude Opus 4.7 · 71.00%

Gap1.2 behindWeightDisplay only
React Native EvalsScore84.7%Versus best verified row

Best verified: Composer 2 · 96.1%

Gap11.4 behindWeightDisplay only
cursorBench31Score59.2%Versus best verified row

Best verified: Claude Fable 5 · 70.6%

Gap11.4 behindWeightDisplay only
FrontierCode 1.1 MainScore43.0%Versus best verified row

Best verified: Claude Fable 5 · 53.5%

Gap10.5 behindWeightDisplay only
SWE-bench (Vals)SWE-bench, Vals AI runScore82.6%Versus best verified row

Best verified: Claude Opus 5 · 97.0%

Gap14.4 behindWeightDisplay only
Agentic13 rows
Agentic benchmark values, best verified comparison, weight, and source status
Terminal-Bench 2.0Score82%Versus best verified row

Best verified: GPT-5.6 Sol · 91.9%

Gap9.9 behindWeightWeighted 30%
BrowseCompScore84.4%Versus best verified row

Best verified: GPT-5.6 Sol · 92.2%

Gap7.8 behindWeightWeighted 25%
OSWorld-VerifiedScore78.7%Versus best verified row

Best verified: Qwen3.8 Max · 86.1%

Gap7.4 behindWeightWeighted 25%
OSWorld 2.0Score13.0%Versus best verified row

Best verified: GPT-6 Astra · 72.6%

Gap59.6 behindWeightWeighted 10%
Benchmark exact
CyberGymScore81.8%Versus best verified row

Best verified: Fugu Cyber · 86.9%

Gap5.1 behindWeightDisplay only
MCP AtlasScore75.3%Versus best verified row

Best verified: Muse Spark 1.1 · 88.1%

Gap12.8 behindWeightDisplay only
ToolathlonScore55.6%Versus best verified row

Best verified: Muse Spark 1.1 · 75.6%

Gap20 behindWeightDisplay only
τ²-bench resultsτ²-Bench Tool-Agent-User EvaluationScore98%Versus best verified row

Best verified: GPT-5.4 · 98.9%

Gap0.9 behindWeightDisplay only
Gert LabsGert Labs Composite Game BenchmarkScore72.93%Versus best verified row

Best verified: Claude Opus 4.8 · 72.97%

Gap0 behindWeightDisplay only
Benchmark exact
ResearchClawBenchScore17.0%Versus best verified row

Best verified: Claude Opus 4.8 · 21.1%

Gap4.1 behindWeightDisplay only
JobBenchScore42.7%Versus best verified row

Best verified: Muse Spark 1.3 · 64.9%

Gap22.2 behindWeightDisplay only
Benchmark exact
ExploitGymScore13.4%Versus best verified row

Best verified: GPT-6 Astra · 42.4%

Gap29 behindWeightDisplay only
Benchmark exact
Terminal-Bench 2.1 (Vals)Terminal-Bench 2.1, Vals AI runScore76.4%Versus best verified row

Best verified: GPT-6 Astra · 87.3%

Gap10.9 behindWeightDisplay only
Reasoning4 rows
Reasoning benchmark values, best verified comparison, weight, and source status
ARC-AGI-2Abstraction and Reasoning Corpus for AGI v2Score85%Versus best verified row

Best verified: GPT-6 Astra · 95%

Gap10 behindWeightWeighted 25%
ARC-AGI-3Abstraction and Reasoning Corpus for AGI v3Score0.4%Versus best verified row

Best verified: GPT-6 Astra · 62.7%

Gap62.3 behindWeightWeighted 15%
MRCR v2 64K-128KOpenAI MRCR v2 8-needle 64K-128KScore83.1%Versus best verified row

Best verified: Gemini 3.7 Flash · 97%

Gap13.9 behindWeightDisplay only
MRCR v2 128K-256KOpenAI MRCR v2 8-needle 128K-256KScore87.5%Versus best verified row

Best verified: GPT-5.5 · 87.5%

GapBest verifiedWeightDisplay only
Knowledge6 rows
Knowledge benchmark values, best verified comparison, weight, and source status
HLEHumanity's Last ExamScore52.2%Versus best verified row

Best verified: Claude Fable 5.1 · 65%

Gap12.8 behindWeightWeighted 35%
HLE w/o toolsHumanity's Last Exam without toolsScore41.4%Versus best verified row

Best verified: Claude Fable 5.1 · 60.9%

Gap19.5 behindWeightWeighted 10%
MMLU-Pro (Vals)MMLU-Pro, Vals AI runScore88.1%Versus best verified row

Best verified: Claude Fable 5.1 · 92.4%

Gap4.3 behindWeightWeighted 10%
GPQAGraduate-Level Google-Proof Q&AScore93.6%Versus best verified row

Best verified: GPT-6 Astra · 96%

Gap2.4 behindWeightWeighted 7%
GPQA-DGPQA DiamondScore93.6%Versus best verified row

Best verified: GPT-6 Astra · 96.0%

Gap2.4 behindWeightDisplay only
GPQA Diamond (Vals)GPQA Diamond, Vals AI runScore93.2%Versus best verified row

Best verified: Gemini 3.1 Pro · 95.5%

Gap2.3 behindWeightDisplay only
Math3 rows
Math benchmark values, best verified comparison, weight, and source status
FrontierMath v2 (Tiers 1-3)FrontierMath v2 Tiers 1-3Score51.700%Versus best verified row

Best verified: GPT-5.6 Sol · 89.000%

Gap37.3 behindWeightWeighted 30%
FrontierMath v2 (Tier 4)FrontierMath v2 Tier 4Score35.400%Versus best verified row

Best verified: GPT-6 Astra · 97.600%

Gap62.2 behindWeightWeighted 10%
FrontierMath (legacy)FrontierMath legacy aggregateScore51.7%Versus best verified row

Best verified: GPT-5.6 Sol · 89%

Gap37.3 behindWeightDisplay only
Multimodal3 rows
Multimodal benchmark values, best verified comparison, weight, and source status
MMMU-ProMassive Multi-discipline Multimodal Understanding ProScore81.2%Versus best verified row

Best verified: Gemini 3.5 Flash · 83.6%

Gap2.4 behindWeightWeighted 40%
OfficeQA ProScore54.1%Versus best verified row

Best verified: Claude Opus 5 · 66.9%

Gap12.8 behindWeightWeighted 25%
MMMU-Pro w/ PythonMMMU-Pro with PythonScore83.2%Versus best verified row

Best verified: GPT-5.6 Sol · 84.6%

Gap1.4 behindWeightDisplay only
External signals4 rows
External signals benchmark values, best verified comparison, weight, and source status
SEC-Bench ProScore45.8%Versus best verified row

Best verified: GPT-6 Astra · 85.4%

Gap39.6 behindWeightDisplay only
Provider exact
Atomic network attacksAtomic Network Attack SimulationScore100%Versus best verified row

Best verified: GPT-6 Astra · 100%

GapBest verifiedWeightDisplay only
Atomic vulnerability researchAtomic Vulnerability Research and ExploitationScore92%Versus best verified row

Best verified: GPT-6 Astra · 100%

Gap8 behindWeightDisplay only
Atomic evasionAtomic EvasionScore54%Versus best verified row

Best verified: GPT-5.6 Sol · 56%

Gap2 behindWeightDisplay only

Lineage

The sequence follows explicit supersedes links. A successor's displayed score stays at least 0.1 points above its predecessor; raw benchmark rows do not move. Scores and prices remain blank when the corresponding public row or first-party rate is unavailable.

  1. Mar 5, 2026

    GPT-5.4

    Score 71.0 · $2.5 / $15

  2. Apr 23, 2026 · you are here

    GPT-5.5

    Score 73.3 · $5 / $30

  3. Jul 9, 2026

    GPT-5.6 Sol

    Score 79.6 · $5 / $30

Base entry

How to read this profile

The visual layer above carries the decisions. These notes preserve the model, ranking, coverage, and family context behind the numbers.

GPT-5.5 ranks #8 of 232 on the public leaderboard with a score of 73.27/100. Its source-verified position is #7 of 130.

GPT-5.5 is a proprietary model with a 1M context window. It uses an explicit reasoning mode, which can improve complex problem solving while adding latency and token use.

Official exact-value snapshot from OpenAI's April 23, 2026 GPT-5.5 launch post, supplemented by CursorBench v3.1 and CursorBench v3.2's highest published GPT-5.5 setting. BenchLM maps only directly comparable public benchmark rows from the launch tables. JobBench uses the Codex CLI GPT-5.5 row reported by the JobBench paper. ExploitGym is normalized from 120 successful exploits out of 898 tasks in the public paper.

GPT-5.5 sits in the GPT-5.5 family with GPT-5.5 Pro. Its explicit predecessor is GPT-5.4. 42 of 422 tracked benchmark slots currently have displayable evidence. Missing categories stay blank.

Its strongest eligible category is Knowledge at #7, while its lowest eligible position is Multimodal & Grounded at #19. particularly effective for knowledge-intensive tasks like research, analysis, and factual Q&A.

Radar

GPT-5.5 release history

Full release history

Radar confirmed these at the source. Already tracking GPT-5.5? See the free Radar Brief for what changes next.

Frequently asked questions

How does GPT-5.5 perform overall in AI benchmarks?

GPT-5.5 ranks #8 out of 232 models on the public BenchAlign leaderboard, with a score of 73.27/100. Its evidence status is Supported, and this profile shows 42 source-displayable benchmark rows. The label describes evidence depth, not a provider quality claim; inspect category rows before choosing a workload.

Is GPT-5.5 good for knowledge and understanding?

GPT-5.5 ranks #7 out of 181 eligible models for knowledge and understanding, with a public category score of 73.3/100. That places it in the current top ten for this category. Check the underlying rows before treating the aggregate as a workload guarantee.

Is GPT-5.5 good for coding and programming?

GPT-5.5 ranks #8 out of 183 eligible models for coding and programming, with a public category score of 67.7/100. That places it in the current top ten for this category. Check the underlying rows before treating the aggregate as a workload guarantee.

Is GPT-5.5 good for mathematics?

GPT-5.5 has source-displayable benchmark coverage for mathematics, but the public category table does not assign it a rank there. The individual rows remain available for inspection. A missing category position means the evidence threshold was not met; it does not convert the model's unmeasured work into a zero.

Is GPT-5.5 good for reasoning and logic?

GPT-5.5 ranks #15 out of 22 eligible models for reasoning and logic, with a public category score of 63.5/100. Higher-ranked alternatives are available for workloads where this category decides the choice. Check the underlying rows before treating the aggregate as a workload guarantee.

Is GPT-5.5 good for agentic tool use and computer tasks?

GPT-5.5 ranks #15 out of 151 eligible models for agentic tool use and computer tasks, with a public category score of 63.9/100. Higher-ranked alternatives are available for workloads where this category decides the choice. Check the underlying rows before treating the aggregate as a workload guarantee.

Is GPT-5.5 good for multimodal and grounded tasks?

GPT-5.5 ranks #19 out of 48 eligible models for multimodal and grounded tasks, with a public category score of 71.3/100. Higher-ranked alternatives are available for workloads where this category decides the choice. Check the underlying rows before treating the aggregate as a workload guarantee.

Which sibling models are related to GPT-5.5?

GPT-5.5 belongs to the GPT-5.5 family. Related tracked variants include GPT-5.5 Pro. A sibling link indicates shared lineage or a documented configuration relationship; it does not mean the variants have identical pricing, context limits, benchmark evidence, or deployment behavior. Compare before switching.

Does GPT-5.5 have full benchmark coverage on BenchLM?

No. GPT-5.5 currently has 62 source-displayable rows across 422 tracked benchmark slots. The profile exposes published, non-generated evidence and leaves missing categories blank until an exact evaluation is available. Coverage describes how much was measured; it is not a penalty added to an individual benchmark result.

What is the context window size of GPT-5.5?

GPT-5.5 has a documented context window of 1M. That figure is the maximum combined prompt and retained-conversation space reported for this exact model; it is not the maximum output length. The profile keeps output limits separate because providers often publish those limits independently.

Compare GPT-5.5 with every tracked model410 comparisons

Last updated September 4, 2026. Runtime fields remain blank until a sourced snapshot exists.

Watch GPT-5.5 in the weekly brief

Get one weekly email when material rank, price, availability, or benchmark evidence changes are worth revisiting.

Read a sample issue

Join 2,000+ readers.