Agentic
Like-for-like- GPT-5.4
- 56.4
- Supported · #35/151
- GPT-5.5
- 63.9
- Supported · #15/151
- Basis
- BenchAlign lane · 13 vs 13 public rows
- Reading
- GPT-5.5 leads · intervals overlap
GPT-5.4 vs GPT-5.5
Updated September 4, 2026. Public scores include evidence status and uncertainty. They are not guarantees for a specific workload.
Decision reading
GPT-5.5 has the higher public score estimate, 73.27 versus 70.96, but the 90% score intervals overlap. Treat that as a lead, not a settled winner.
25 results are shared. Category rows resting on Estimated evidence or different benchmark sets are marked directional and do not name a winner.
Share or export
Which one for your work
Recommendations appear only when a shared evidence basis or an explicit operating constraint supports the call. Secondary and unsupported use cases stay disclosed below the initial list.
Coding work
Code generation, repair, and software-engineering tasks
GPT-5.5
GPT-5.5 leads on the public coding lane, 67.7 to 54, with Supported evidence for both models, although the 90% intervals overlap.
Confidence: limited
Agentic work
Tool use, computer use, and multi-step task completion
GPT-5.5
GPT-5.5 leads on the public agentic lane, 63.9 to 56.4, with Supported evidence for both models, although the 90% intervals overlap.
Confidence: limited
Long documents
Prompts that approach the documented context limit
GPT-5.4
GPT-5.4 has the larger documented context window.
Confidence: documented
Show secondary and unsupported calls
Chat turn cost
1K fresh input + 500 output tokens
GPT-5.4
GPT-5.4 has the lower estimated token cost for this stated workload. Costs use the listed standard API rates.
Confidence: listed-rates
Cache-heavy agent loop cost
200K cached + 20K fresh input + 10K output tokens
GPT-5.4
GPT-5.4 has the lower estimated token cost for this stated workload. Costs use the listed standard API rates.
Confidence: listed-rates
Repository review cost
50K fresh input + 3K output tokens
GPT-5.4
GPT-5.4 has the lower estimated token cost for this stated workload. Costs use the listed standard API rates.
Confidence: listed-rates
What is actually comparable
Shared results can support a head-to-head reading. Results present for only one model describe coverage, not superiority.
12 GPT-5.4 only
25 Shared
13 GPT-5.5 only
- Shared results
- 25
- GPT-5.4 only
- 12
- GPT-5.5 only
- 13
- Like-for-like categories
- 4 / 8
2 categories rest on Estimated evidence or different evidence sets. Those rows remain visible for coverage context but do not name a winner.
Category results, on a stated basis
Each row shows the public-lane category score for both models: the BenchAlign lane for agentic, coding, and knowledge, and the weighted public rows elsewhere, with the same rank each model page reports. A row is like-for-like only when both scores rest on Supported evidence or the same weighted set. Directional and not-comparable rows remain visible, but they never receive a winner.
Coding
Like-for-like- GPT-5.4
- 54.0
- Supported · #49/183
- GPT-5.5
- 67.7
- Supported · #8/183
- Basis
- BenchAlign lane · 4 vs 9 public rows
- Reading
- GPT-5.5 leads · intervals overlap
Reasoning
Like-for-like- GPT-5.4
- 57.0
- #19/22
- GPT-5.5
- 63.5
- #15/22
- Basis
- Provisional lane · 2 vs 2 weighted rows
- Reading
- GPT-5.5 leads
Knowledge
Like-for-like- GPT-5.4
- 69.2
- Supported · #16/181
- GPT-5.5
- 73.3
- Supported · #7/181
- Basis
- BenchAlign lane · 7 vs 6 public rows
- Reading
- GPT-5.5 leads · intervals overlap
Multimodal
Directional only- GPT-5.4
- 69.3
- #20/48
- GPT-5.5
- 71.3
- #19/48
- Basis
- Provisional lane · 3 vs 2 weighted rows
- Reading
- Directional only
Instruction following
Directional only- GPT-5.4
- 90.4
- #19/120
- GPT-5.5
- 92.9
- #7/120
- Basis
- Provisional lane · 0 vs 0 weighted rows
- Reading
- Directional only
Math
Not comparable- GPT-5.4
- 64.5
- Unranked · 2 rankable rows
- GPT-5.5
- 69.6
- Unranked · 3 rankable rows
- Basis
- Provisional lane · 2 vs 2 weighted rows
- Reading
- Not comparable
Multilingual
Not comparable- GPT-5.4
- Not ranked
- GPT-5.5
- Not ranked
- Basis
- Provisional lane · 0 vs 0 weighted rows
- Reading
- Not comparable
| Category | GPT-5.4 | GPT-5.5 | Basis | Reading |
|---|---|---|---|---|
| Agentic | 56.4Supported · #35/151 | 63.9Supported · #15/151 | Like-for-likeBenchAlign lane · 13 vs 13 public rows | GPT-5.5 leads · intervals overlap |
| Coding | 54.0Supported · #49/183 | 67.7Supported · #8/183 | Like-for-likeBenchAlign lane · 4 vs 9 public rows | GPT-5.5 leads · intervals overlap |
| Reasoning | 57.0#19/22 | 63.5#15/22 | Like-for-likeProvisional lane · 2 vs 2 weighted rows | GPT-5.5 leads |
| Knowledge | 69.2Supported · #16/181 | 73.3Supported · #7/181 | Like-for-likeBenchAlign lane · 7 vs 6 public rows | GPT-5.5 leads · intervals overlap |
| Multimodal | 69.3#20/48 | 71.3#19/48 | Directional onlyProvisional lane · 3 vs 2 weighted rows | Directional only |
| Instruction following | 90.4#19/120 | 92.9#7/120 | Directional onlyProvisional lane · 0 vs 0 weighted rows | Directional only |
| Math | 64.5Unranked · 2 rankable rows | 69.6Unranked · 3 rankable rows | Not comparableProvisional lane · 2 vs 2 weighted rows | Not comparable |
| Multilingual | Not ranked | Not ranked | Not comparableProvisional lane · 0 vs 0 weighted rows | Not comparable |
Ranks count the models scored in each category’s lane, so the agentic, coding, and knowledge denominators (BenchAlign) differ from the provisional-lane categories. Unranked scores sit on the lane’s scale but never name a winner. Like-for-like rows within 0.5 points read as a practical tie.
Shape of the matched evidence
Only shared public evidence is shown. Sparse evidence stays a ruled list rather than being closed into a radar shape.
Too few matched category axes support a radar. The ruled list below shows only shared benchmark results; positions use each benchmark’s normalized display scale when available.
ARC-AGI-2
Reasoning
FrontierMath v2 (Tier 4)
Math
Terminal-Bench 2.0
Agentic
FrontierMath v2 (Tiers 1-3)
Math
OSWorld-Verified
Agentic
What each workload costs
Three fixed token mixes turn per-token rates into comparable decisions. Each scenario states context fit and whether cached input had to fall back to the published list-input rate.
Chat turn
1K fresh input + 500 output tokens
- GPT-5.4
- $0.01
- Fits in one request
- GPT-5.5
- $0.02
- Fits in one request
GPT-5.4 has the lower modeled cost
Costs use the listed standard API rates.
Repository review
50K fresh input + 3K output tokens
- GPT-5.4
- $0.17
- Fits in one request
- GPT-5.5
- $0.34
- Fits in one request
GPT-5.4 has the lower modeled cost
Costs use the listed standard API rates.
Cache-heavy agent loop
200K cached + 20K fresh input + 10K output tokens
- GPT-5.4
- $0.25
- Fits in one request
- GPT-5.5
- $0.5
- Fits in one request
GPT-5.4 has the lower modeled cost
Costs use the listed standard API rates.
Specification differences
Sourced differences are shown directly. Missing facts stay explicit instead of being inferred from a model name or family.
SpecificationGPT-5.4GPT-5.5
Context window
Maximum documented context; output-token limits may be lower.
GPT-5.4
1.05M
OpenAI pricingGPT-5.5
API model ID
GPT-5.4
gpt-5.4
OpenAI pricingGPT-5.5
gpt-5.5
OpenAI pricingCached-input rate
A missing cached-input rate falls back to the listed input rate only in the stated workload estimate.
GPT-5.4
$0.25 per 1M cached input tokens
OpenAI pricingGPT-5.5
$0.5 per 1M cached input tokens
OpenAI pricingDocumented inputs
GPT-5.4
Not sourced
GPT-5.5
Not sourced
Documented outputs
GPT-5.4
Not sourced
GPT-5.5
Not sourced
Provider availability
GPT-5.4
Not sourced
GPT-5.5
Not sourced
Reasoning profile
GPT-5.4
Reasoning
GPT-5.5
Reasoning
Weight access
GPT-5.4
Proprietary
GPT-5.5
Proprietary
License
GPT-5.4
Proprietary
GPT-5.5
Proprietary
Release date
GPT-5.4
2026-03-05
GPT-5.5
2026-04-23
If you are considering the documented upgrade path
- Deployment change
- Both entries list OpenAI as the provider. Confirm endpoint, model ID, limits, and feature support before switching.
- Quality signal
- GPT-5.5 has the higher public score estimate, 73.27 versus 70.96, but the 90% score intervals overlap.
- Workload cost
- Repository review: $0.17 vs $0.34. Cache-heavy agent loop: $0.25 vs $0.5.
- Context tradeoff
- GPT-5.4 has the larger documented window (1.05M).
Run the same representative tasks against both endpoints before changing production traffic.
Benchmark evidence
The full public result ledger is available for audit without forcing a wide desktop table onto a phone.
Browse raw public benchmark evidence50 rows
Agentic
Terminal-Bench 2.0
GPT-5.475.1%SourceGPT-5.582%SourceGPT-5.5 leads this result
CyberGym
Shared sourceGPT-5.479.0%GPT-5.581.8%GPT-5.5 leads this result
BrowseComp
GPT-5.482.7%SourceGPT-5.584.4%SourceGPT-5.5 leads this result
OSWorld-Verified
GPT-5.475%SourceGPT-5.578.7%SourceGPT-5.5 leads this result
MCP Atlas
Shared sourceGPT-5.470.6%GPT-5.575.3%GPT-5.5 leads this result
Toolathlon
GPT-5.454.6%SourceGPT-5.555.6%SourceGPT-5.5 leads this result
τ²-bench results
GPT-5.498.9%SourceGPT-5.598%SourceGPT-5.4 leads this result
Claw-Eval
GPT-5.460.3%SourceGPT-5.5—Not directly comparable
DeepSearchQA
GPT-5.473.6%SourceGPT-5.5—Not directly comparable
Gert Labs
Shared sourceGPT-5.464.89%GPT-5.572.93%GPT-5.5 leads this result
ResearchClawBench
Shared sourceGPT-5.415.3%GPT-5.517.0%GPT-5.5 leads this result
JobBench
Shared sourceGPT-5.438.9%GPT-5.542.7%GPT-5.5 leads this result
ExploitGym
Shared sourceGPT-5.46.0%GPT-5.513.4%GPT-5.5 leads this result
OSWorld 2.0
GPT-5.4—GPT-5.513.0%SourceNot directly comparable
Terminal-Bench 2.1 (Vals)
GPT-5.4—GPT-5.576.4%SourceNot directly comparable
Coding
LiveCodeBench Pro
GPT-5.487.5%SourceGPT-5.5—Not directly comparable
SWE-bench Pro
GPT-5.457.7%SourceGPT-5.558.6%SourceGPT-5.5 leads this result
React Native Evals
Shared sourceGPT-5.485.3%GPT-5.584.7%GPT-5.4 leads this result
Vibe Code Bench
Shared sourceGPT-5.467.42%GPT-5.569.85%GPT-5.5 leads this result
Terminal-Bench 2.0
GPT-5.4—GPT-5.582.0%SourceNot directly comparable
cursorBench31
GPT-5.4—GPT-5.559.2%SourceNot directly comparable
cursorBench32
GPT-5.4—GPT-5.558.4%SourceNot directly comparable
FrontierCode 1.1 Main
GPT-5.4—GPT-5.543.0%SourceNot directly comparable
LiveCodeBench (Vals)
GPT-5.4—GPT-5.585.3%SourceNot directly comparable
SWE-bench (Vals)
GPT-5.4—GPT-5.582.6%SourceNot directly comparable
Reasoning
ARC-AGI-2
GPT-5.474.0%SourceGPT-5.585%SourceGPT-5.5 leads this result
ARC-AGI-3
Shared sourceGPT-5.40.2%GPT-5.50.4%GPT-5.5 leads this result
MRCR v2 64K-128K
GPT-5.4—GPT-5.583.1%SourceNot directly comparable
MRCR v2 128K-256K
GPT-5.4—GPT-5.587.5%SourceNot directly comparable
Knowledge
GPQA
GPT-5.492.8%SourceGPT-5.593.6%SourceGPT-5.5 leads this result
- GPT-5.452.1%GPT-5.552.2%
GPT-5.5 leads this result
HLE w/o tools
GPT-5.439.8%SourceGPT-5.541.4%SourceGPT-5.5 leads this result
GPQA-D
GPT-5.492.8%SourceGPT-5.593.6%SourceGPT-5.5 leads this result
HealthBench Hard
GPT-5.440.1%SourceGPT-5.5—Not directly comparable
MedXpertQA (Text)
GPT-5.459.6%SourceGPT-5.5—Not directly comparable
HealthBench Professional
GPT-5.448.1%SourceGPT-5.5—Not directly comparable
GPQA Diamond (Vals)
GPT-5.4—GPT-5.593.2%SourceNot directly comparable
MMLU-Pro (Vals)
GPT-5.4—GPT-5.588.1%SourceNot directly comparable
Math
FrontierMath v2 (Tiers 1-3)
Shared sourceGPT-5.447.600%GPT-5.551.700%GPT-5.5 leads this result
FrontierMath v2 (Tier 4)
Shared sourceGPT-5.427.100%GPT-5.535.400%GPT-5.5 leads this result
FrontierMath (legacy)
GPT-5.4—GPT-5.551.7%SourceNot directly comparable
Multimodal
MMMU-Pro
GPT-5.481.2%SourceGPT-5.581.2%SourceTie
OfficeQA Pro
Shared sourceGPT-5.453.2%GPT-5.554.1%GPT-5.5 leads this result
MMMU-Pro w/ Python
Shared sourceGPT-5.482.1%GPT-5.583.2%GPT-5.5 leads this result
CharXiv
GPT-5.482.8%SourceGPT-5.5—Not directly comparable
ERQA
GPT-5.465.4%SourceGPT-5.5—Not directly comparable
SimpleVQA
GPT-5.461.1%SourceGPT-5.5—Not directly comparable
ScreenSpot Pro
GPT-5.485.4%SourceGPT-5.5—Not directly comparable
ZeroBench
GPT-5.441.0%SourceGPT-5.5—Not directly comparable
MedXpertQA (MM)
GPT-5.477.1%SourceGPT-5.5—Not directly comparable
Frequently asked questions
Which is better, GPT-5.4 or GPT-5.5?
GPT-5.5 has the higher public score estimate, 73.27 versus 70.96, but the 90% score intervals overlap. The higher estimate is not a decisive winner because the uncertainty ranges overlap.
Which is better for coding, GPT-5.4 or GPT-5.5?
GPT-5.5 leads the public coding lane, 67.7 to 54, with Supported evidence for both models, although the 90% intervals overlap.
Which is better for agentic tasks, GPT-5.4 or GPT-5.5?
GPT-5.5 leads the public agentic tasks lane, 63.9 to 56.4, with Supported evidence for both models, although the 90% intervals overlap.
Which costs less, GPT-5.4 or GPT-5.5?
For the stated presets, chat costs $0.01 on GPT-5.4 and $0.02 on GPT-5.5; repository review costs $0.17 and $0.34; the cache-heavy agent loop costs $0.25 and $0.5. Costs use the listed standard API rates.
Which has the larger context window, GPT-5.4 or GPT-5.5?
GPT-5.4 has the larger documented context window: 1.05M, compared with 1M.
Related comparisons
Last updated September 4, 2026
Watch GPT-5.4 vs GPT-5.5
One weekly email when material rank, price, or benchmark evidence changes make this matchup worth revisiting.
Read a sample issueJoin 2,000+ readers.