Model comparison
GPT-5.4 vs GPT-5.5
Head-to-head evidence from 40 shared benchmark results across 7 categories. Overall scores shown here use the public BenchAlign v5 ranking lane.
Public leaderboard positions: GPT-5.4 #8 (Supported); GPT-5.5 #9 (Estimated). Intervals and evidence labels describe ranking uncertainty, not a guarantee for a specific workload.
Evidence parity. GPT-5.4 and GPT-5.5 share 40 comparable benchmark results. 5 of 8 categories are comparable. 12 results are unique to GPT-5.4; 17 to GPT-5.5.
Updated July 23, 2026- Shared results
- 40
- GPT-5.4 only
- 12
- GPT-5.5 only
- 17
- Comparable categories
- 5 / 8
Pick GPT-5.4 if you want the stronger benchmark profile. GPT-5.5 only becomes the better choice if mathematics is the priority.
Confidence note. This is a partial-evidence comparison with 40 shared benchmark results across 7 evidence categories; 5 of 8 categories currently have scoreable aggregates for both models. Treat the verdict as directional until coverage is more balanced.
Why this result
GPT-5.4 has the cleaner BenchAlign overall profile here, landing at 74.24 versus 73.51. It is a real lead, but still close enough that category-level strengths matter more than the headline number.
GPT-5.4's sharpest advantage is in multimodal & grounded, where it averages 73.2 against 70.4. The single biggest benchmark swing on the page is FrontierMath v2 (Tier 4), 27.100% to 35.400%. GPT-5.5 does hit back in mathematics, so the answer changes if that is the part of the workload you care about most.
GPT-5.5 is also the more expensive model on tokens at $5.00 input / $30.00 output per 1M tokens, versus $2.50 input / $15.00 output per 1M tokens for GPT-5.4. That is roughly 2.0x on output cost alone. GPT-5.4 gives you the larger context window at 1.05M, compared with 1M for GPT-5.5.
Category breakdown
Exact category averages are shown below. Not measured means BenchLM does not have enough sourced public coverage for that model and category.
| Category | GPT-5.4 | Δ | GPT-5.5 |
|---|---|---|---|
| Math | GPT-5.442.5 | Margin→ 5.1 | GPT-5.547.6 |
| Agentic | GPT-5.477.2 | Margin→ 4.4 | GPT-5.581.6 |
| Multimodal | GPT-5.473.2 | Margin← 2.8 | GPT-5.570.4 |
| Coding | GPT-5.457.7 | Margin→ 0.9 | GPT-5.558.6 |
| Knowledge | GPT-5.457.6 | Margin→ 0.2 | GPT-5.557.8 |
| Reasoning | GPT-5.4Not measured | MarginNo overlap | GPT-5.585.0 |
Decisive benchmark drivers
The largest measured benchmark gaps in this matchup, with exact reported values.
More
- Source ↗
FrontierMath v2 (Tier 4)
MathA 27.100%B 35.400%Winner: GPT-5.5Δ 8.3FrontierMath v2 (Tier 4): GPT-5.4 scored 27.100%; GPT-5.5 scored 35.400%. GPT-5.5 wins this benchmark. - Source ↗
Terminal-Bench 2.0
AgenticA 75.1%B 82%Winner: GPT-5.5Δ 6.9Terminal-Bench 2.0: GPT-5.4 scored 75.1%; GPT-5.5 scored 82%. GPT-5.5 wins this benchmark. - Source ↗
FrontierMath v2 (Tiers 1-3)
MathA 47.600%B 51.700%Winner: GPT-5.5Δ 4.1FrontierMath v2 (Tiers 1-3): GPT-5.4 scored 47.600%; GPT-5.5 scored 51.700%. GPT-5.5 wins this benchmark. - Source ↗
OSWorld-Verified
AgenticA 75%B 78.7%Winner: GPT-5.5Δ 3.7OSWorld-Verified: GPT-5.4 scored 75%; GPT-5.5 scored 78.7%. GPT-5.5 wins this benchmark. - Source ↗
BrowseComp
AgenticA 82.7%B 84.4%Winner: GPT-5.5Δ 1.7BrowseComp: GPT-5.4 scored 82.7%; GPT-5.5 scored 84.4%. GPT-5.5 wins this benchmark.
Operational comparison
Runtime and commercial metrics are compared only when both models have a complete sourced value.
| Metric | GPT-5.4 | GPT-5.5 | Comparison |
|---|---|---|---|
| Input / output priceUSD per 1M tokens | GPT-5.4$2.5 input / $15 output | GPT-5.5$5 input / $30 output | GPT-5.4 has the lower combined listed price. |
| Generation speedtokens per second | GPT-5.474 tok/s | GPT-5.5Not available | A complete speed comparison is not available. |
| First-answer latencyseconds to first token | GPT-5.4151.79 s | GPT-5.5Not available | A complete latency comparison is not available. |
| Context windowmaximum listed tokens | GPT-5.41.05M | GPT-5.51M | GPT-5.4 lists the larger context window. |
Benchmark Deep Dive
AgenticGPT-5.5 wins26 benchmarks
| Benchmark | GPT-5.4 | GPT-5.5 | Result |
|---|---|---|---|
| Terminal-Bench 2.0Source | 75.1% | 82% | GPT-5.5 leads |
| CyberGymSource | 79.0% | 81.8% | GPT-5.5 leads |
| BrowseCompSource | 82.7% | 84.4% | GPT-5.5 leads |
| OSWorld-VerifiedSource | 75% | 78.7% | GPT-5.5 leads |
| MCP AtlasSource | 70.6% | 75.3% | GPT-5.5 leads |
| ToolathlonSource | 54.6% | 55.6% | GPT-5.5 leads |
| τ²-bench resultsSource | 87.1% | 93.9% | GPT-5.5 leads |
| Claw-EvalSource | 60.3% | — | Not comparable |
| DeepSearchQASource | 73.6% | — | Not comparable |
| AA Agentic IndexSource | 41.1% | 44.9% | GPT-5.5 leads |
| APEX-Agents-AASource | 33.3% | 37.7% | GPT-5.5 leads |
| GDPval-AASource | 44.7% | 49.5% | GPT-5.5 leads |
| GDPval-AASource | 1395 | 1490 | GPT-5.5 leads |
| Gert LabsSource | 64.89% | 72.93% | GPT-5.5 leads |
| ResearchClawBenchSource | 15.3% | 17.0% | GPT-5.5 leads |
| JobBenchSource | 38.9% | 42.7% | GPT-5.5 leads |
| ExploitGymSource | 6.0% | 13.4% | GPT-5.5 leads |
| OSWorld 2.0Source | — | 13.0% | Not comparable |
| AA BriefcaseSource | — | 1154 | Not comparable |
| AA AutomationBenchSource | — | 42.1% | Not comparable |
| AA EnterpriseOps-GymSource | — | 46.6% | Not comparable |
| AA Harvey LABSource | — | 86.3% | Not comparable |
| AA ITBenchSource | — | 45.8% | Not comparable |
| AA Tau3 BankingSource | — | 31.3% | Not comparable |
| terminalBenchHardSource | — | 60.6% | Not comparable |
| aaTerminalBench21Source | — | 84.3% | Not comparable |
CodingGPT-5.5 wins10 benchmarks
| Benchmark | GPT-5.4 | GPT-5.5 | Result |
|---|---|---|---|
| LiveCodeBench ProSource | 87.5% | — | Not comparable |
| SWE-bench ProSource | 57.7% | 58.6% | GPT-5.5 leads |
| React Native EvalsSource | 85.3% | 84.7% | GPT-5.4 leads |
| Vibe Code BenchSource | 67.42% | 69.85% | GPT-5.5 leads |
| AA Coding IndexSource | 71.0% | 74.9% | GPT-5.5 leads |
| AA-SciCodeSource | 56.6% | 56.1% | GPT-5.4 leads |
| Terminal-Bench 2.0Source | — | 82.0% | Not comparable |
| cursorBench31Source | — | 59.2% | Not comparable |
| cursorBench32Source | — | 58.4% | Not comparable |
| FrontierCode 1.1 MainSource | — | 43.0% | Not comparable |
Reasoning5 benchmarks
KnowledgeGPT-5.5 wins13 benchmarks
| Benchmark | GPT-5.4 | GPT-5.5 | Result |
|---|---|---|---|
| GPQASource | 92.8% | 93.6% | GPT-5.5 leads |
| HLESource | 52.1% | 52.2% | GPT-5.5 leads |
| HLE w/o toolsSource | 39.8% | 41.4% | GPT-5.5 leads |
| GPQA-DSource | 92.8% | 93.6% | GPT-5.5 leads |
| HealthBench HardSource | 40.1% | — | Not comparable |
| MedXpertQA (Text)Source | 59.6% | — | Not comparable |
| Artificial Analysis Intelligence IndexSource | 51.4% | 54.8% | GPT-5.5 leads |
| AA-GPQA DiamondSource | 92.0% | 93.5% | GPT-5.5 leads |
| AA-HLESource | 41.6% | 44.3% | GPT-5.5 leads |
| AA-Omniscience IndexSource | 5.7% | 20.1% | GPT-5.5 leads |
| AA-Omniscience AccuracySource | 50.0% | 56.9% | GPT-5.5 leads |
| AA-Omniscience Hallucination RateSource | 88.6% | 85.5% | GPT-5.5 leads |
| HealthBench ProfessionalSource | 48.1% | — | Not comparable |
MathGPT-5.5 wins3 benchmarks
MultimodalGPT-5.4 wins11 benchmarks
| Benchmark | GPT-5.4 | GPT-5.5 | Result |
|---|---|---|---|
| MMMU-ProSource | 81.2% | 81.2% | Tie |
| OfficeQA ProSource | 53.2% | 54.1% | GPT-5.5 leads |
| MMMU-Pro w/ PythonSource | 82.1% | 83.2% | GPT-5.5 leads |
| CharXivSource | 82.8% | — | Not comparable |
| ERQASource | 65.4% | — | Not comparable |
| SimpleVQASource | 61.1% | — | Not comparable |
| ScreenSpot ProSource | 85.4% | — | Not comparable |
| ZeroBenchSource | 41.0% | — | Not comparable |
| MedXpertQA (MM)Source | 77.1% | — | Not comparable |
| AA-MMMU-ProSource | 78.4% | 79.9% | GPT-5.5 leads |
| Design Arena WebsiteSource | 1250 | 1282 | GPT-5.5 leads |
Inst. Following1 benchmarks
| Benchmark | GPT-5.4 | GPT-5.5 | Result |
|---|---|---|---|
| AA-IFBenchSource | 73.9% | 75.9% | GPT-5.5 leads |
Frequently Asked Questions (6)
Which is better, GPT-5.4 or GPT-5.5?
GPT-5.4 is ahead on BenchLM's BenchAlign leaderboard, 74.24 to 73.51. The biggest single separator in this matchup is FrontierMath v2 (Tier 4), where the scores are 27.100% and 35.400%.
Which is better for knowledge tasks, GPT-5.4 or GPT-5.5?
GPT-5.5 has the edge for knowledge tasks in this comparison, averaging 57.8 versus 57.6. Inside this category, AA-Omniscience Index is the benchmark that creates the most daylight between them.
Which is better for coding, GPT-5.4 or GPT-5.5?
GPT-5.5 has the edge for coding in this comparison, averaging 58.6 versus 57.7. Inside this category, AA Coding Index is the benchmark that creates the most daylight between them.
Which is better for math, GPT-5.4 or GPT-5.5?
GPT-5.5 has the edge for math in this comparison, averaging 47.6 versus 42.5. Inside this category, FrontierMath v2 (Tier 4) is the benchmark that creates the most daylight between them.
Which is better for agentic tasks, GPT-5.4 or GPT-5.5?
GPT-5.5 has the edge for agentic tasks in this comparison, averaging 81.6 versus 77.2. Inside this category, GDPval-AA is the benchmark that creates the most daylight between them.
Which is better for multimodal and grounded tasks, GPT-5.4 or GPT-5.5?
GPT-5.4 has the edge for multimodal and grounded tasks in this comparison, averaging 73.2 versus 70.4. Inside this category, Design Arena Website is the benchmark that creates the most daylight between them.
Related Comparisons
Explore More
Choose a model with this week’s evidence
Join 2,000+ readers for ranking moves, pricing changes, and the claims that still need proof.
One email each week. Unsubscribe anytime.