LLM Reference

GPT-5.5

Released
2026-04-23
Last refreshed
2026-06-29
Status
Researched 51d ago
ProprietaryCommercial use: conditionalMultimodalCodingRAGAgentsLong contextVisionClassificationJSON / Tool use

GPT-5.5 is worth evaluating for coding, rag, and agents when its provider route and context window match the workload.

Use it for

  • Teams evaluating coding, rag, and agents
  • Workloads that can use a 1.05m context window
  • Buyers comparing 4 tracked provider routes

Do not use it for

  • Workloads where another current model has stronger sourced task evidence
Specifications
Family
GPT-5.5
Released
2026-04-23
Context
1.05m
Max output
128,000
Architecture
Decoder Only
Knowledge cutoff
2025-12
Specialization
general
Openness
Proprietary
License
ProprietaryCommercial use: conditional
Weights
Not released
Code
Unknown
Created by
OpenAIOpenAI

Cutting-edge research and development.

San Francisco, California, United States
Founded 2015
Website
Pricing
Output / 1M
$30.00
Input / 1M
$5.00

Cheapest of 4 routes · OpenAI API · cache read $0.500

Providers(4)
OpenAI APIOpenRouterVercel AI GatewayAWS Bedrock
View 4 provider routes
Links
Website

About

GPT-5.5 is OpenAI's fully retrained agentic model, released April 23, 2026. Optimised for agentic coding, computer use, knowledge work, and early scientific research. Achieves 82.7% on Terminal-Bench 2.0 (Codex CLI scaffold), 84.9% on GDPval, 58.6% on SWE-Bench Pro, 93.6% on GPQA Diamond, and 82.6% on SWE-Bench Verified (Vals.ai independent harness). Knowledge cutoff December 2025. Supports reasoning effort levels (none/low/medium/high/xhigh). Context window 1,050,000 tokens with a long-context surcharge above 272K tokens. Model ID: gpt-5.5.

GPT-5.5 is a proprietary model. The structured metadata tracks a 1.05m-token context window, multimodal input, reasoning, function calling, tool use, structured outputs, and code execution. This page tracks provider routes through OpenAI API, OpenRouter, Vercel AI Gateway, and 1 more, with the cheapest tracked route listed at $5 input and $30 output per 1M tokens. Headline tracked benchmarks include Google-Proof Q&A 93.6, SWE-bench Pro 58.6, and Chatbot Arena 1488.0.

Top use-case fit: coding, agents, and build tasks

Coding

Q/$ D

4 relevant benchmarks in the decision map.

RAG

Included by capability and metadata signals in the decision map.

Agents

Q/$ D

1 relevant benchmark in the decision map.

Provider price ladder

Compare all 4

Compare API pricing across 4 providers for input and output tokens, batch, and cached reads when available.

ProviderInput / 1MOutput / 1MBatch in / outCacheRoute
OpenAI API$5.00$30.00$2.50 / $15.00read $0.500
Serverless
OpenRouter$5.00$30.00--
Serverless
Vercel AI Gateway$5.00$30.00-read $0.500
Serverless
AWS Bedrock$5.50$33.00--
Serverless

Available via routers & gateways(16)

LiteLLM

Gateway

Open-source Python SDK and proxy server that unifies 100+ LLM APIs behind a single OpenAI-compatible interface, with load balancing, cost tracking, and configurable failover.

Free OSSOpenAI API

OpenRouter

Hybrid

Unified hybrid gateway to 400+ models from 60+ providers via a single OpenAI-compatible API, with optional auto-routing that selects the best model per prompt.

PassthroughOpenAI API

Portkey

Gateway

Production AI gateway routing to 1,600+ LLMs with failover, load balancing, semantic caching, and guardrails; Apache 2.0 core is fully self-hostable with the complete feature set.

SubscriptionOpenAI API

AIRouter

Router

Commercial LLM router that analyzes incoming requests and routes to the optimal model for cost/quality/latency via a drop-in OpenAI-compatible API, with a privacy-preserving embedding mode that avoids sending prompt content.

Passthrough + feeOpenAI API

Amazon Bedrock Intelligent Prompt Routing

Router

AWS Bedrock's native intelligent prompt router that routes prompts between Anthropic Claude model tiers (Haiku/Sonnet) based on predicted task complexity, with no extra per-routing charge.

PassthroughAWS Bedrock

Helicone

Gateway

Observability-first AI gateway with routing, caching, rate limiting, and request tracing; Apache 2.0 open-source core with a managed hosted tier for logging and analytics.

SubscriptionOpenAI API

Capabilities

VisionMultimodalReasoningFunction CallingTool UseStructured OutputsCode ExecutionPrompt CachingBatch API

Benchmark peer barsfor Coding

SWE-bench ProRank 13 of 42
Claude Fable 5
80.3
Claude Opus 5
79.2
Claude Opus 4.8
69.2
Grok 4.5
64.7
GPT-5.5current
58.6
SWE-bench VerifiedRank 8 of 81
Claude Fable 5
96.0
Claude Opus 5
96.0
Claude Mythos Preview
93.9
Claude Opus 4.8
88.6
GPT-5.5current
82.6
HumanEvalRank 5 of 97
Claude Sonnet 4.6
98.0
o3
96.7
Claude Opus 4.6
95.0
Grok-3
94.5
GPT-5.5current
94.2

Benchmark scores(30)

Scores are benchmark-specific and are direction-aware: the same numeric gap can mean very different outcomes across suites. Use the leaderboard context and this model's provider route to decide whether the winning margin is meaningful for your workload.
BenchmarkScoreVersionEvaluationSource
Google-Proof Q&A93.6diamondObserved 2026-04-23—Source
SWE-bench Pro58.6SWE-bench Pro (pass@1)Observed 2026-06-07—Source
Chatbot Arena1488.0HighObserved 2026-04-28—Source
MMLU PRO88.1—Observed 2026-04-29—Source
MMMU Pro88.3Vals.ai standardized CoT harnessObserved 2026-06-02—Source
HumanEval94.2—Observed 2026-04-29—Source
Instruction-Following Evaluation92.1—Observed 2026-04-29—Source
SWE-bench Verified82.6SWE-bench VerifiedObserved 2026-04-30—Source
Massive Multitask Language Understanding92.4MMLU (accuracy)Observed 2026-06-07—Source
Terminal-Bench 2.082.7Terminal-Bench 2.0 (accuracy%)Observed 2026-06-07—Source
Aider Polyglot88.0Listed as 'gpt-5 (high)' = 88 (percent_correct)Observed 2026-06-07—Source
AIME 202581.2AIME 2025 (accuracy)Observed 2026-06-07—Source
ARC-AGI-285.0ARC-AGI-2 (accuracy%)Observed 2026-06-07—Source
BrowseComp84.4BrowseComp (accuracy%)Observed 2026-06-07—Source
Humanity's Last Exam41.4HLE without tools (accuracy)Observed 2026-06-07—Source
MCP-Atlas75.3MCP-Atlas (accuracy%)Observed 2026-06-07—Source
GDPval84.9GDPval official launch scoreObserved 2026-04-23—Source
Terminal-Bench 2.178.2Terminal-Bench 2.1Observed 2026-06-18—Source
OSWorld-Verified78.7OSWorld-VerifiedObserved 2026-06-09—Source
GDPval-AA1769.0GDPval-AA ELOObserved 2026-06-09—Source
Blueprint-Bench 236.2Blueprint-Bench 2Observed 2026-06-09—Source
AutomationBench12.9AutomationBenchObserved 2026-06-09—Source
Legal Agent Benchmark2.1Legal Agent BenchmarkObserved 2026-06-09—Source
GDP.pdf24.9GDP.pdf vision, no toolsObserved 2026-06-09—Source
CursorBench64.3CursorBench 3.1Observed 2026-06-30
Configuration: GPT-5.5 Extra High
Harness: CursorBench 3.1
Evaluator: Cursor
Confidence: confirmed
Notes: Highest CursorBench 3.1 score across Cursor's published effort configurations for this base model.
Source
GeneBench-Pro12.0xhighObserved 2026-06-30—Source
CursorBench58.4CursorBench 3.2Observed 2026-07-18
Configuration: GPT-5.5 High
Harness: CursorBench 3.2 productized Cursor-agent workflow
Evaluator: Cursor
Confidence: confirmed
Cost/task: $2.05
Tokens/task: 12,183
Steps/task: 28
Notes: Cursor vendor-reported result; not independently reproducible. Results are subject to variance, and small score differences may not be statistically meaningful.
Source
CursorBench58.4CursorBench 3.2Observed 2026-07-18
Configuration: GPT-5.5 Extra High
Harness: CursorBench 3.2 productized Cursor-agent workflow
Evaluator: Cursor
Confidence: confirmed
Cost/task: $2.85
Tokens/task: 17,534
Steps/task: 32
Notes: Cursor vendor-reported result; not independently reproducible. Results are subject to variance, and small score differences may not be statistically meaningful.
Source
CursorBench53.8CursorBench 3.2Observed 2026-07-18
Configuration: GPT-5.5 Medium
Harness: CursorBench 3.2 productized Cursor-agent workflow
Evaluator: Cursor
Confidence: confirmed
Cost/task: $1.51
Tokens/task: 8,522
Steps/task: 25
Notes: Cursor vendor-reported result; not independently reproducible. Results are subject to variance, and small score differences may not be statistically meaningful.
Source
CursorBench46.6CursorBench 3.2Observed 2026-07-18
Configuration: GPT-5.5 Low
Harness: CursorBench 3.2 productized Cursor-agent workflow
Evaluator: Cursor
Confidence: confirmed
Cost/task: $0.98
Tokens/task: 5,168
Steps/task: 20
Notes: Cursor vendor-reported result; not independently reproducible. Results are subject to variance, and small score differences may not be statistically meaningful.
Source

Migration checks

Coming from GPT-4o?

Compare GPT-5.5 with other models

Show all 78 popular comparisonssorted by 7-day search impressions
GPT-5.5 vs GPT-5.22KGPT-5.5 vs GLM-5.11KGPT-5.5 vs Grok 41KGPT-5.5 vs DeepSeek V4 Pro1KGPT-5.5 vs o3 Mini900GPT-5.5 vs Claude Opus 4.5825GPT-5.5 vs Gemini 3.5 Flash755GPT-5.5 vs Claude Sonnet 4.5695GPT-5.5 vs GPT-5.3-Codex687GPT-5.5 vs Claude Mythos Preview537GPT-5.5 vs o3 Deep Research510GPT-5.5 vs Gemini Deep Research Preview400GPT-5.5 vs GLM-5397GPT-5.5 vs Qwen3.6-27B341GPT-5.5 vs DeepSeek V4 Flash340GPT-5.5 vs Claude Opus 4.6338GPT-5.5 vs GPT-5.2 Codex303GPT-5.5 vs Qwen3.6-35B-A3B212GPT-5.5 vs DeepSeek R1193GPT-5.5 vs Grok-3174GPT-5.5 vs Gemini 2.5 Flash169GPT-5.5 vs Qwen2.5-72B-Instruct166GPT-5.5 vs Kimi K2.5162GPT-5.5 vs Gemini 2.5 Pro159GPT-5.5 vs Mistral Large 2140GPT-5.5 vs Qwen3-Max131GPT-5.5 vs DeepSeek R1 0528124GPT-5.5 vs Trinity-Large-Thinking116GPT-5.5 vs Gemini 2.5 Pro Preview 06-05114GPT-5.5 vs Llama 3.1 70B Instruct103GPT-5.5 vs GPT-5.5 Pro96GPT-5.5 vs Claude 3.7 Sonnet89GPT-5.5 vs GPT-5.4 Pro81GPT-5.5 vs Gemini 3.1 Flash TTS Preview77GPT-5.5 vs GPT-574GPT-5.5 vs Mistral Large 3 675B Instruct71GPT-5.5 vs Qwen2.5-72B68GPT-5.5 vs DeepSeek V363GPT-5.5 vs Qwen3.5-27B59GPT-5.5 vs Qwen3.5-35B-A3B58GPT-5.5 vs Llama 3.2 1B54GPT-5.5 vs DeepSeek V3.251GPT-5.5 vs Phi-3 Mini 4k49GPT-5.5 vs Llama 3.2 1B Instruct41GPT-5.5 vs Gemini 3.1 Pro Preview Custom Tools41GPT-5.5 vs Mixtral 8x7B38GPT-5.5 vs Gemma 7B Instruct37GPT-5.5 vs Llama 3 8B Instruct36GPT-5.5 vs Code Davinci 00136GPT-5.5 vs Kimi K2 Thinking Turbo35GPT-5.5 vs Llama 3 70B Instruct35GPT-5.5 vs Grok 3 Mini33GPT-5.5 vs Qwen3.6 Max Preview30GPT-5.5 vs Phi-4 Mini Flash Reasoning30GPT-5.5 vs Mixtral 8x22B Instruct v0.327GPT-5.5 vs Qwen3.7-Max25GPT-5.5 vs Qwen3.5-122B-A10B23GPT-5.5 vs Together AI - Llama 3 8B Lite18GPT-5.5 vs Mistral Nemotron17GPT-5.5 vs Gemini 2.5 Pro Preview 05-0616GPT-5.5 vs Nano Banana (Gemini 2.5 Flash Image)16GPT-5.5 vs Gemini 2.5 Flash Live API15GPT-5.5 vs Qwen2-7B-Instruct13GPT-5.5 vs Together AI Qwen2-7B-Instruct11GPT-5.5 vs Qwen3.5-397B-A17B10GPT-5.5 vs Phi-4 Reasoning Vision 15B10GPT-5.5 vs Gemma 2 9B SahabatAI Instruct8GPT-5.5 vs DeepSeek R1 Lite7GPT-5.5 vs DeepSeek R1 Basic7GPT-5.5 vs Qwen3.6-Max6GPT-5.5 vs Qwen3-235B-A22B6GPT-5.5 vs Gemini 3.1 Flash Lite Preview6GPT-5.5 vs DeepSeek V3 Base1GPT-5.5 vs ShieldGemma 9B1GPT-5.5 vs Llama 2 13B Chat1GPT-5.5 vs GLM-5V-Turbo1GPT-5.5 vs Qwen3.5-9B1GPT-5.5 vs Gemini 3.5 Pro0

Frequently asked questions

What is the context window of GPT-5.5?

GPT-5.5 has a context window of 1.05m tokens.

What is the max output of GPT-5.5?

GPT-5.5 can generate up to 128,000 output tokens.

How much does GPT-5.5 cost?

GPT-5.5 pricing ranges from $5/1M to $5.50/1M input tokens depending on the provider.

When was GPT-5.5 released?

GPT-5.5 was released on 2026-04-23.

Which providers offer GPT-5.5?

GPT-5.5 is available from 4 providers: OpenAI API, OpenRouter, Vercel AI Gateway, AWS Bedrock.

What benchmarks has GPT-5.5 been tested on?

GPT-5.5 has been evaluated on 30 benchmarks, including Google-Proof Q&A, SWE-bench Pro, Chatbot Arena, MMLU PRO, MMMU Pro.