Model comparison
Claude Sonnet 5 vs GPT-5.6 Terra
Head-to-head evidence from 23 shared benchmark results across 5 categories. Overall scores shown here use the public BenchAlign v5 ranking lane.
Public leaderboard positions: Claude Sonnet 5 #29 (Estimated); GPT-5.6 Terra #11 (Estimated). Intervals and evidence labels describe ranking uncertainty, not a guarantee for a specific workload.
Evidence parity. Claude Sonnet 5 and GPT-5.6 Terra share 23 comparable benchmark results. 4 of 8 categories are comparable. 13 results are unique to Claude Sonnet 5; 21 to GPT-5.6 Terra.
Updated July 23, 2026- Shared results
- 23
- Claude Sonnet 5 only
- 13
- GPT-5.6 Terra only
- 21
- Comparable categories
- 4 / 8
Pick GPT-5.6 Terra if you want the stronger benchmark profile. Claude Sonnet 5 only becomes the better choice if coding is the priority or you want the cheaper token bill.
Confidence note. This is a partial-evidence comparison with 23 shared benchmark results across 5 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 65.32. 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 57.4. The single biggest benchmark swing on the page is Terminal-Bench 2.0, 80.4% to 87.4%. Claude Sonnet 5 does hit back in coding, so the answer changes if that is the part of the workload you care about most.
GPT-5.6 Terra is also the more expensive model on tokens at $2.50 input / $15.00 output per 1M tokens, versus $2.00 input / $10.00 output per 1M tokens for Claude Sonnet 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 | Claude Sonnet 5 | Δ | GPT-5.6 Terra |
|---|---|---|---|
| Knowledge | Claude Sonnet 557.4 | Margin→ 35.5 | GPT-5.6 Terra92.9 |
| Coding | Claude Sonnet 576.7 | Margin← 13.3 | GPT-5.6 Terra63.4 |
| Multimodal | Claude Sonnet 588.3 | Margin← 7.6 | GPT-5.6 Terra80.7 |
| Agentic | Claude Sonnet 581.9 | Margin→ 5.5 | GPT-5.6 Terra87.4 |
| Math | Claude Sonnet 5Not measured | MarginNo overlap | GPT-5.6 Terra80.8 |
Decisive benchmark drivers
The largest measured benchmark gaps in this matchup, with exact reported values.
More
- Source ↗
Terminal-Bench 2.0
AgenticA 80.4%B 87.4%Winner: GPT-5.6 TerraΔ 7Terminal-Bench 2.0: Claude Sonnet 5 scored 80.4%; GPT-5.6 Terra scored 87.4%. GPT-5.6 Terra wins this benchmark. - Source ↗
BrowseComp
AgenticA 84.7%B 87.5%Winner: GPT-5.6 TerraΔ 2.8BrowseComp: Claude Sonnet 5 scored 84.7%; GPT-5.6 Terra scored 87.5%. GPT-5.6 Terra wins this benchmark. - Source ↗
SWE-bench Pro
CodingA 63.2%B 63.4%Winner: GPT-5.6 TerraΔ 0.2SWE-bench Pro: Claude Sonnet 5 scored 63.2%; GPT-5.6 Terra scored 63.4%. GPT-5.6 Terra wins this benchmark.
Operational comparison
Runtime and commercial metrics are compared only when both models have a complete sourced value.
| Metric | Claude Sonnet 5 | GPT-5.6 Terra | Comparison |
|---|---|---|---|
| Input / output priceUSD per 1M tokens | Claude Sonnet 5$2 input / $10 output | GPT-5.6 Terra$2.5 input / $15 output | Claude Sonnet 5 has the lower combined listed price. |
| Generation speedtokens per second | Claude Sonnet 5Not available | GPT-5.6 TerraNot available | A complete speed comparison is not available. |
| First-answer latencyseconds to first token | Claude Sonnet 5Not available | GPT-5.6 TerraNot available | A complete latency comparison is not available. |
| Context windowmaximum listed tokens | Claude Sonnet 51M | GPT-5.6 Terra1M | Listed context windows are equal. |
Benchmark Deep Dive
AgenticGPT-5.6 Terra wins21 benchmarks
| Benchmark | Claude Sonnet 5 | GPT-5.6 Terra | Result |
|---|---|---|---|
| Terminal-Bench 2.0Source | 80.4% | 87.4% | GPT-5.6 Terra leads |
| BrowseCompSource | 84.7% | 87.5% | GPT-5.6 Terra leads |
| HLE w/ toolsSource | 57.4% | — | Not comparable |
| OSWorld-VerifiedSource | 81.2% | — | Not comparable |
| GDPval-AASource | 1607 | 1581 | Claude Sonnet 5 leads |
| AA Agentic IndexSource | 46.7% | 47.4% | GPT-5.6 Terra leads |
| GDPval-AASource | 55.4% | 54.1% | Claude Sonnet 5 leads |
| AA BriefcaseSource | 1388 | — | Not comparable |
| AA AutomationBenchSource | 39.2% | 45.6% | GPT-5.6 Terra leads |
| AA EnterpriseOps-GymSource | 44.7% | — | Not comparable |
| AA Harvey LABSource | 90.1% | 85.2% | Claude Sonnet 5 leads |
| AA Tau3 BankingSource | 28.2% | 31.8% | GPT-5.6 Terra leads |
| aaTerminalBench21Source | 80.5% | 88% | GPT-5.6 Terra leads |
| OSWorld 2.0Source | — | 50.2% | Not comparable |
| CyberGymSource | — | 81.8% | Not comparable |
| ExploitGymSource | — | 23.2% | Not comparable |
| ToolathlonSource | — | 53.1% | Not comparable |
| τ²-bench resultsSource | — | 86.3% | Not comparable |
| AA ITBenchSource | — | 51.0% | Not comparable |
| terminalBenchHardSource | — | 57.6% | Not comparable |
| APEX-Agents-AASource | — | 38.9% | Not comparable |
CodingClaude Sonnet 5 wins11 benchmarks
| Benchmark | Claude Sonnet 5 | GPT-5.6 Terra | Result |
|---|---|---|---|
| SWE-bench VerifiedSource | 85.2% | — | Not comparable |
| SWE-bench ProSource | 63.2% | 63.4% | GPT-5.6 Terra leads |
| SWE MultilingualSource | 78.3% | — | Not comparable |
| SWE MultimodalSource | 28.1% | — | Not comparable |
| Terminal-Bench 2.0Source | 80.4% | 87.4% | GPT-5.6 Terra leads |
| FrontierCode 1.1 MainSource | 42.7% | — | Not comparable |
| cursorBench32Source | 61.5% | 64.9% | GPT-5.6 Terra leads |
| AA Coding IndexSource | 71.5% | 76.7% | GPT-5.6 Terra leads |
| AA-SciCodeSource | 53.6% | 53.9% | GPT-5.6 Terra leads |
| deepSweSource | — | 69.6% | Not comparable |
| FrontierCode 1.1 ExtendedSource | — | 55.8% | Not comparable |
Reasoning3 benchmarks
KnowledgeGPT-5.6 Terra wins12 benchmarks
| Benchmark | Claude Sonnet 5 | GPT-5.6 Terra | Result |
|---|---|---|---|
| HLESource | 57.4% | — | Not comparable |
| HLE w/o toolsSource | 43.2% | — | Not comparable |
| Artificial Analysis Intelligence IndexSource | 53.4% | 55.0% | GPT-5.6 Terra leads |
| AA-GPQA DiamondSource | 91.1% | 92.5% | GPT-5.6 Terra leads |
| AA-HLESource | 39.6% | 41.8% | GPT-5.6 Terra leads |
| AA-Omniscience IndexSource | 15.3% | -0.2% | Claude Sonnet 5 leads |
| AA-Omniscience AccuracySource | 38.3% | 45.9% | GPT-5.6 Terra leads |
| AA-Omniscience Hallucination RateSource | 37.3% | 85.2% | Claude Sonnet 5 leads |
| GPQASource | — | 92.9% | Not comparable |
| GPQA-DSource | — | 92.9% | Not comparable |
| HealthBench ProfessionalSource | — | 57.7% | Not comparable |
| HealthBench HardSource | — | 32.7% | Not comparable |
Math3 benchmarks
MultimodalClaude Sonnet 5 wins6 benchmarks
Inst. Following1 benchmarks
| Benchmark | Claude Sonnet 5 | GPT-5.6 Terra | Result |
|---|---|---|---|
| AA-IFBenchSource | — | 71.2% | Not comparable |
Frequently Asked Questions (5)
Which is better, Claude Sonnet 5 or GPT-5.6 Terra?
GPT-5.6 Terra is ahead on BenchLM's BenchAlign leaderboard, 72.57 to 65.32. The biggest single separator in this matchup is Terminal-Bench 2.0, where the scores are 80.4% and 87.4%.
Which is better for knowledge tasks, Claude Sonnet 5 or GPT-5.6 Terra?
GPT-5.6 Terra has the edge for knowledge tasks in this comparison, averaging 92.9 versus 57.4. Inside this category, AA-Omniscience Hallucination Rate is the benchmark that creates the most daylight between them.
Which is better for coding, Claude Sonnet 5 or GPT-5.6 Terra?
Claude Sonnet 5 has the edge for coding in this comparison, averaging 76.7 versus 63.4. Inside this category, Terminal-Bench 2.0 is the benchmark that creates the most daylight between them.
Which is better for agentic tasks, Claude Sonnet 5 or GPT-5.6 Terra?
GPT-5.6 Terra has the edge for agentic tasks in this comparison, averaging 87.4 versus 81.9. Inside this category, GDPval-AA is the benchmark that creates the most daylight between them.
Which is better for multimodal and grounded tasks, Claude Sonnet 5 or GPT-5.6 Terra?
Claude Sonnet 5 has the edge for multimodal and grounded tasks in this comparison, averaging 88.3 versus 80.7. Inside this category, AA-MMMU-Pro is the benchmark that creates the most daylight between them.
Related Comparisons
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.