Model comparison
GPT-5.4 Pro vs GPT-5.6 Terra
Head-to-head evidence from 6 shared benchmark results across 4 categories. Overall scores shown here use the public BenchAlign v5 ranking lane.
Public leaderboard positions: GPT-5.4 Pro #44 (Estimated); GPT-5.6 Terra #11 (Estimated). Intervals and evidence labels describe ranking uncertainty, not a guarantee for a specific workload.
Evidence parity. GPT-5.4 Pro and GPT-5.6 Terra share 6 comparable benchmark results. 4 of 8 categories are comparable. 6 results are unique to GPT-5.4 Pro; 38 to GPT-5.6 Terra.
Updated July 23, 2026- Shared results
- 6
- GPT-5.4 Pro only
- 6
- GPT-5.6 Terra only
- 38
- Comparable categories
- 4 / 8
Pick GPT-5.6 Terra if you want the stronger benchmark profile. GPT-5.4 Pro only becomes the better choice if multimodal & grounded is the priority or you need the larger 1.05M context window.
Confidence note. This is a partial-evidence comparison with 6 shared benchmark results across 4 evidence categories; 4 of 8 categories currently have scoreable aggregates for both models. Treat the verdict as directional until coverage is more balanced.
Why this result
GPT-5.6 Terra is clearly ahead on the BenchAlign aggregate, 72.57 to 60.89. The gap is large enough that you do not need to squint at the spreadsheet to see the difference.
GPT-5.6 Terra's sharpest advantage is in knowledge, where it averages 92.9 against 58.7. The single biggest benchmark swing on the page is FrontierMath v2 (Tiers 1-3), 50.000% to 84.900%. GPT-5.4 Pro does hit back in multimodal & grounded, so the answer changes if that is the part of the workload you care about most.
GPT-5.4 Pro is also the more expensive model on tokens at $30.00 input / $180.00 output per 1M tokens, versus $2.50 input / $15.00 output per 1M tokens for GPT-5.6 Terra. That is roughly 12.0x on output cost alone. GPT-5.4 Pro gives you the larger context window at 1.05M, compared with 1M for GPT-5.6 Terra.
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 Pro | Δ | GPT-5.6 Terra |
|---|---|---|---|
| Knowledge | GPT-5.4 Pro58.7 | Margin→ 34.2 | GPT-5.6 Terra92.9 |
| Math | GPT-5.4 Pro46.9 | Margin→ 33.9 | GPT-5.6 Terra80.8 |
| Multimodal | GPT-5.4 Pro94.0 | Margin← 13.3 | GPT-5.6 Terra80.7 |
| Agentic | GPT-5.4 Pro89.3 | Margin← 1.9 | GPT-5.6 Terra87.4 |
| Coding | GPT-5.4 ProNot measured | MarginNo overlap | GPT-5.6 Terra63.4 |
| Reasoning | GPT-5.4 Pro83.3 | MarginNo overlap | GPT-5.6 TerraNot measured |
Decisive benchmark drivers
The largest measured benchmark gaps in this matchup, with exact reported values.
More
- Source ↗
FrontierMath v2 (Tiers 1-3)
MathA 50.000%B 84.900%Winner: GPT-5.6 TerraΔ 34.9FrontierMath v2 (Tiers 1-3): GPT-5.4 Pro scored 50.000%; GPT-5.6 Terra scored 84.900%. GPT-5.6 Terra wins this benchmark. - Source ↗
FrontierMath v2 (Tier 4)
MathA 37.500%B 68.300%Winner: GPT-5.6 TerraΔ 30.8FrontierMath v2 (Tier 4): GPT-5.4 Pro scored 37.500%; GPT-5.6 Terra scored 68.300%. GPT-5.6 Terra wins this benchmark. - Source ↗
MMMU-Pro
MultimodalA 94%B 80.7%Winner: GPT-5.4 ProΔ 13.3MMMU-Pro: GPT-5.4 Pro scored 94%; GPT-5.6 Terra scored 80.7%. GPT-5.4 Pro wins this benchmark. - Source ↗
BrowseComp
AgenticA 89.3%B 87.5%Winner: GPT-5.4 ProΔ 1.8BrowseComp: GPT-5.4 Pro scored 89.3%; GPT-5.6 Terra scored 87.5%. GPT-5.4 Pro wins this benchmark.
Operational comparison
Runtime and commercial metrics are compared only when both models have a complete sourced value.
| Metric | GPT-5.4 Pro | GPT-5.6 Terra | Comparison |
|---|---|---|---|
| Input / output priceUSD per 1M tokens | GPT-5.4 Pro$30 input / $180 output | GPT-5.6 Terra$2.5 input / $15 output | GPT-5.6 Terra has the lower combined listed price. |
| Generation speedtokens per second | GPT-5.4 Pro74 tok/s | GPT-5.6 TerraNot available | A complete speed comparison is not available. |
| First-answer latencyseconds to first token | GPT-5.4 Pro151.79 s | GPT-5.6 TerraNot available | A complete latency comparison is not available. |
| Context windowmaximum listed tokens | GPT-5.4 Pro1.05M | GPT-5.6 Terra1M | GPT-5.4 Pro lists the larger context window. |
Benchmark Deep Dive
AgenticGPT-5.4 Pro wins17 benchmarks
| Benchmark | GPT-5.4 Pro | GPT-5.6 Terra | Result |
|---|---|---|---|
| BrowseCompSource | 89.3% | 87.5% | GPT-5.4 Pro leads |
| Terminal-Bench 2.0Source | — | 87.4% | Not comparable |
| OSWorld 2.0Source | — | 50.2% | Not comparable |
| CyberGymSource | — | 81.8% | Not comparable |
| ExploitGymSource | — | 23.2% | Not comparable |
| ToolathlonSource | — | 53.1% | Not comparable |
| AA Agentic IndexSource | — | 47.4% | Not comparable |
| τ²-bench resultsSource | — | 86.3% | Not comparable |
| GDPval-AASource | — | 54.1% | Not comparable |
| GDPval-AASource | — | 1581 | Not comparable |
| AA Harvey LABSource | — | 85.2% | Not comparable |
| AA ITBenchSource | — | 51.0% | Not comparable |
| AA Tau3 BankingSource | — | 31.8% | Not comparable |
| AA AutomationBenchSource | — | 45.6% | Not comparable |
| terminalBenchHardSource | — | 57.6% | Not comparable |
| aaTerminalBench21Source | — | 88% | Not comparable |
| APEX-Agents-AASource | — | 38.9% | Not comparable |
Coding7 benchmarks
| Benchmark | GPT-5.4 Pro | GPT-5.6 Terra | Result |
|---|---|---|---|
| SWE-bench ProSource | — | 63.4% | Not comparable |
| Terminal-Bench 2.0Source | — | 87.4% | Not comparable |
| deepSweSource | — | 69.6% | Not comparable |
| FrontierCode 1.1 ExtendedSource | — | 55.8% | Not comparable |
| cursorBench32Source | — | 64.9% | Not comparable |
| AA Coding IndexSource | — | 76.7% | Not comparable |
| AA-SciCodeSource | — | 53.9% | Not comparable |
Reasoning4 benchmarks
KnowledgeGPT-5.6 Terra wins14 benchmarks
| Benchmark | GPT-5.4 Pro | GPT-5.6 Terra | Result |
|---|---|---|---|
| HLESource | 58.7% | — | Not comparable |
| FrontierScienceSource | 36.7% | — | Not comparable |
| FrontierScience ResearchSource | 36.7% | — | Not comparable |
| HLE w/o toolsSource | 42.7% | — | Not comparable |
| GPQASource | — | 92.9% | Not comparable |
| GPQA-DSource | — | 92.9% | Not comparable |
| HealthBench ProfessionalSource | — | 57.7% | Not comparable |
| HealthBench HardSource | — | 32.7% | Not comparable |
| Artificial Analysis Intelligence IndexSource | — | 55.0% | Not comparable |
| AA-GPQA DiamondSource | — | 92.5% | Not comparable |
| AA-HLESource | — | 41.8% | Not comparable |
| AA-Omniscience IndexSource | — | -0.2% | Not comparable |
| AA-Omniscience AccuracySource | — | 45.9% | Not comparable |
| AA-Omniscience Hallucination RateSource | — | 85.2% | Not comparable |
MathGPT-5.6 Terra wins4 benchmarks
MultimodalGPT-5.4 Pro wins3 benchmarks
Inst. Following1 benchmarks
| Benchmark | GPT-5.4 Pro | GPT-5.6 Terra | Result |
|---|---|---|---|
| AA-IFBenchSource | — | 71.2% | Not comparable |
Frequently Asked Questions (5)
Which is better, GPT-5.4 Pro or GPT-5.6 Terra?
GPT-5.6 Terra is ahead on BenchLM's BenchAlign leaderboard, 72.57 to 60.89. The biggest single separator in this matchup is FrontierMath v2 (Tiers 1-3), where the scores are 50.000% and 84.900%.
Which is better for knowledge tasks, GPT-5.4 Pro or GPT-5.6 Terra?
GPT-5.6 Terra has the edge for knowledge tasks in this comparison, averaging 92.9 versus 58.7. GPT-5.4 Pro stays close enough that the answer can still flip depending on your workload.
Which is better for math, GPT-5.4 Pro or GPT-5.6 Terra?
GPT-5.6 Terra has the edge for math in this comparison, averaging 80.8 versus 46.9. Inside this category, FrontierMath (legacy) is the benchmark that creates the most daylight between them.
Which is better for agentic tasks, GPT-5.4 Pro or GPT-5.6 Terra?
GPT-5.4 Pro has the edge for agentic tasks in this comparison, averaging 89.3 versus 87.4. Inside this category, BrowseComp is the benchmark that creates the most daylight between them.
Which is better for multimodal and grounded tasks, GPT-5.4 Pro or GPT-5.6 Terra?
GPT-5.4 Pro has the edge for multimodal and grounded tasks in this comparison, averaging 94 versus 80.7. Inside this category, MMMU-Pro 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.