GPT-5.6 Terra vs GPT-6 Sol
| Spec | GPT-5.6 Terra | GPT-6 Sol |
|---|---|---|
| Respan Index | 55.3 (#28) | 63.1 (#18) |
| Blended price per 1M | $4.5 | $4 |
| Context window | 1.1M | 1.1M |
| Input / output per 1M | $2 / $12 | $2 / $10 |
| Max output | 128K | 128K |
| Input | text, vision | text, vision |
| Released | 2026-07-09 | 2026-09-22 |
Respan Index by category
Each category is scored 0 to 100, and how far ahead the winner is. Leads under a point are close to even.
| Category | GPT-5.6 Terra | GPT-6 Sol |
|---|---|---|
| Coding | 37.0 GPT-5.6 Terra | 44.8 GPT-6 Sol |
| Agents | 63.1 GPT-5.6 Terra | 71.0 GPT-6 Sol |
| Reasoning | 70.5 GPT-5.6 Terra | 77.3 GPT-6 Sol |
| Knowledge & language | 44.7 GPT-5.6 Terra | 60.6 GPT-6 Sol |
| Human preference | 54.5 GPT-5.6 Terra | 58.3 GPT-6 Sol |
Benchmarks both report
GPT-6 Sol leads on 19 of the 26 benchmarks both report. Higher is better. Each score links to where it was published.
| Benchmark | GPT-5.6 Terra | GPT-6 Sol |
|---|---|---|
| LMArena Agent (LMArena) | -0.0267 | 0.0647 |
| LMArena Elo | 1468 | 1458 |
| LMArena WebDev (LMArena) | 1520 | 1674 |
| LMArena Vision (LMArena) | 1266 | 1258 |
| MirrorCode (Epoch AI) | 41.1% | 48.9% |
| Chess Puzzles (Epoch AI) | 47% | 56% |
| APEX-Agents (Mercor) | 58.2% | 54.3% |
| FrontierMath Tier 4 v2 (Epoch AI) | 70.7% | 90% |
| FrontierMath Tiers 1-3 v2 (Epoch AI) | 86% | 89.8% |
| Furniture Assembly (Epoch AI) | 54.2% | 58.3% |
| GPQA Diamond | 92.9% | 94.3% |
| LiveBench Agentic Coding (LiveBench) | 54.9% | 52.9% |
| LiveBench Coding (LiveBench) | 78.2% | 81.8% |
| LiveBench Data Analysis (LiveBench) | 79.3% | 81.2% |
| LiveBench Instruction Following (LiveBench) | 64.6% | 68.6% |
| LiveBench Language (LiveBench) | 82.9% | 85.3% |
| LiveBench Mathematics (LiveBench) | 94.9% | 96.4% |
| LiveBench Reasoning (LiveBench) | 90.6% | 88.7% |
| Mystery Game Puzzles (Epoch AI) | 35% | 56% |
| OTIS Mock AIME 2024-2025 (Epoch AI) | 99.7% | 100% |
| SAGE (Vals AI) | 47% | 44.8% |
| SimpleQA Verified | 43.2% | 60.7% |
| Terminal-Bench 4.0 (Vals AI) | 22.7% | 44.4% |
| Vending-Bench 2 (Andon Labs) | 7343.21 | 14427.85 |
| DeepSWE v1.1 | 69.6% | 68.8% |
| SimpleBench (SimpleBench) | 48.9% | 73.1% |
FAQ
Is GPT-5.6 Terra better than GPT-6 Sol?
GPT-6 Sol scores 63.1 on the Respan Index (#18), ahead of GPT-5.6 Terra at 55.3 (#28). GPT-6 Sol leads on 19 of the 26 benchmarks both report.
Which is cheaper, GPT-5.6 Terra or GPT-6 Sol?
GPT-5.6 Terra lists at $2 input and $12 output per 1M tokens; GPT-6 Sol at $2 and $10. GPT-6 Sol is 1.1x cheaper per token at list price.
Which has the larger context window?
Both accept 1.1M tokens of context.