GPT-5.6 Terra vs Qwen2.5 72B Instruct
GPT-5.6 Terra leads Qwen2.5 72B Instruct on the Artificial Analysis Intelligence Index, 42.1 to 7.7, and Qwen2.5 72B Instruct costs 30.0x less per output token ($0.40 vs $12.00 per million). Updated October 22, 2026; 40 benchmarks have a result for both.
vs boards won of 40 shared
GPT-5.6 Terra or Qwen2.5 72B Instruct: which to pick
Only GPT-5.6 Terra has a coding score, 63.9 on the Vibe Coding Index; Qwen2.5 72B Instruct has none, so the two cannot be ranked on coding.
Pick Qwen2.5 72B Instruct for output tokens 30.0x cheaper ($0.40 vs $12.00 per million).
Model AGPT-5.6 TerraOpenAI63.9/1006 independent benchmarks$0.42 per solved taskModel BQwen2.5 72B InstructQwenNot benchmarkedNo coding benchmarkNo cost per solved task
- Shared results
- 40
- GPT-5.6 Terra only
- 87
- Qwen2.5 72B Instruct only
- 13
- Like-for-like categories
- 7 of 9
Only benchmarks that tested both models count toward a pick; a result only one model has is shown but not compared.
Which one for your work
Each card is decided by one number both models carry. Where only one has it, the card says so and makes no call.
Coding
Vibe Coding Index, six public coding benchmarks
No call
Only GPT-5.6 Terra has a coding score, so no call is made.
One side not measured
Intelligence
AA Intelligence Index
GPT-5.6 Terra
GPT-5.6 Terra leads, 42.1 against 7.7.
Same board, both measured
Agentic work
AA Agentic Index
No call
Only GPT-5.6 Terra has a AA Agentic Index result, 43.2.
Directional only
Reasoning
GPQA Diamond
No call
Only GPT-5.6 Terra has a GPQA Diamond result, 90.9%.
Directional only
Knowledge
MMLU Pro
No call
Only GPT-5.6 Terra has a MMLU Pro result, 86.7%.
Directional only
Web design
Design Arena: websites
No call
Only GPT-5.6 Terra has a Design Arena: websites result, 1268.
Directional only
Long documents
Listed context window
GPT-5.6 Terra
GPT-5.6 Terra takes 1.05M tokens against 33K.
Listed by OpenRouter
Everyday cost
1K input + 500 output tokens
Qwen2.5 72B Instruct
GPT-5.6 Terra $0.0080, Qwen2.5 72B Instruct $0.0006 per turn.
List prices
Cost per solved task
One SWE-bench Verified task, solved
No call
Only GPT-5.6 Terra has the number, $0.42.
Directional only
At a glance
- Coding index
- GPT-5.6 Terra 63.9 from 6 benchmarks, Qwen2.5 72B Instruct not benchmarked.
- Widest category
- Math, 51.1 points apart on average over 3 shared benchmarks.
- Widest single gap
- OTIS Mock AIME: GPT-5.6 Terra 99.7% against Qwen2.5 72B Instruct 8.1%, 91.7 points apart on a 0 to 100 scale.
- Output price
- GPT-5.6 Terra $12.00 against Qwen2.5 72B Instruct $0.40 per million output tokens, 30.0x apart.
- Context window
- GPT-5.6 Terra takes 1.05M tokens, Qwen2.5 72B Instruct 33K.
Benchmark by benchmark
First the score: boards each model leads, per category, where both have a result. Then every board, a category to a fold; bars are each result on a 0 to 100 scale, so compare the two in a row, not one row with another.
| Category | GPT-5.6 Terra | Bars | Qwen2.5 72B Instruct | Ties |
|---|---|---|---|---|
| Coding | 4 | 0 | ||
| Intelligence | 2 | 0 | ||
| Agentic work | 2 | 0 | ||
| Reasoning | 6 | 0 | ||
| Math | 3 | 0 | ||
| Knowledge | 4 | 0 | ||
| General | 19 | 0 | ||
| All shared boards | 40 | 0 |
CodingLike-for-likeSix of these boards feed the Vibe Coding Index40
| Benchmark | GPT-5.6 Terra | Qwen2.5 72B Instruct | Leader |
|---|---|---|---|
| LiveBench coding | 86.2% | Not measured | Only GPT-5.6 Terra measured |
| LMArena WebDev | 1522 | Not measured | Only GPT-5.6 Terra measured |
| AA Coding Index | 76.7 | Not measured | Only GPT-5.6 Terra measured |
| LiveCodeBench (Vals AI) | 85.9% | Not measured | Only GPT-5.6 Terra measured |
| SWE-bench Verified (Vals AI) | 95.4% | Not measured | Only GPT-5.6 Terra measured |
| Terminal-Bench 4 (Vals AI) | 22.7% | Not measured | Only GPT-5.6 Terra measured |
| Vibe Code Bench (Vals AI) | 74.6% | Not measured | Only GPT-5.6 Terra measured |
| Code migration (Vals AI) | 47.8% | Not measured | Only GPT-5.6 Terra measured |
| ProgramBench (Vals AI) | 0.5% | Not measured | Only GPT-5.6 Terra measured |
| IOI (Vals AI) | 87.6% | Not measured | Only GPT-5.6 Terra measured |
| Terminal-Bench 2.1 (Vals AI) | 77.5% | Not measured | Only GPT-5.6 Terra measured |
| Vibe Code Bench 1-100 (Vals AI) | 14.8% | Not measured | Only GPT-5.6 Terra measured |
| SkillsBench (Vals AI) | 58.9% | Not measured | Only GPT-5.6 Terra measured |
| CyberBench (Vals AI) | 72.1% | Not measured | Only GPT-5.6 Terra measured |
| LiveCodeBench (Artificial Analysis run) | Not measured | 27.6% | Only Qwen2.5 72B Instruct measured |
| SciCode | 55.0% | Not measured | Only GPT-5.6 Terra measured |
| Terminal-Bench Hard (Artificial Analysis run) | 62.9% | 4.5% | GPT-5.6 Terra |
| FrontierCode | 41.3% | Not measured | Only GPT-5.6 Terra measured |
| WeirdML | 78.3% | 16.0% | GPT-5.6 Terra |
| ALE-Bench | 1951.4 | Not measured | Only GPT-5.6 Terra measured |
| LMArena Text: coding | 1485 | 1292 | GPT-5.6 Terra |
| LMArena Text: software and IT | 1482 | 1292 | GPT-5.6 Terra |
| LMArena WebDev: React | 1513 | Not measured | Only GPT-5.6 Terra measured |
| LMArena WebDev: HTML | 1526 | Not measured | Only GPT-5.6 Terra measured |
| LMArena WebDev: from an image | 1541 | Not measured | Only GPT-5.6 Terra measured |
| SciCode | 10.8% | Not measured | Only GPT-5.6 Terra measured |
| Neither model has a result on the other 13 boards here | |||
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index20
| Benchmark | GPT-5.6 Terra | Qwen2.5 72B Instruct | Leader |
|---|---|---|---|
| AA Intelligence Index | 42.1 | 7.7 | GPT-5.6 Terra |
| Epoch Capabilities Index | 160.1 | 129.0 | GPT-5.6 Terra |
Agentic workLike-for-likePublic boards, shown as published, kept out of the index20
| Benchmark | GPT-5.6 Terra | Qwen2.5 72B Instruct | Leader |
|---|---|---|---|
| AA Agentic Index | 43.2 | Not measured | Only GPT-5.6 Terra measured |
| Design Arena: agents building web apps | 1120 | Not measured | Only GPT-5.6 Terra measured |
| Design Arena: agents building full-stack apps | 1061 | Not measured | Only GPT-5.6 Terra measured |
| Design Arena: agents building mobile apps | 1111 | Not measured | Only GPT-5.6 Terra measured |
| Design Arena: agents building Android apps | 1061 | Not measured | Only GPT-5.6 Terra measured |
| Design Arena: agents building games | 1083 | Not measured | Only GPT-5.6 Terra measured |
| Design Arena: HTML slides | 1057 | Not measured | Only GPT-5.6 Terra measured |
| Design Arena: PowerPoint slides | 1107 | Not measured | Only GPT-5.6 Terra measured |
| LiveBench research | 60.4% | Not measured | Only GPT-5.6 Terra measured |
| LiveBench browser | 84.0% | Not measured | Only GPT-5.6 Terra measured |
| LiveBench long agentic | 48.4% | Not measured | Only GPT-5.6 Terra measured |
| SREBench (Vals AI) | 9.5% | Not measured | Only GPT-5.6 Terra measured |
| Finance Agent (Vals AI) | 54.4% | Not measured | Only GPT-5.6 Terra measured |
| Tax Agent Bench (Vals AI) | 30.0% | Not measured | Only GPT-5.6 Terra measured |
| Legal Research Bench (Vals AI) | 41.4% | Not measured | Only GPT-5.6 Terra measured |
| Excel Modeling Benchmark (Vals AI) | 66.2% | Not measured | Only GPT-5.6 Terra measured |
| Terminal-Bench-Science (Vals AI) | 10.0% | Not measured | Only GPT-5.6 Terra measured |
| τ²-Bench Telecom | 86.3% | 34.5% | GPT-5.6 Terra |
| APEX-Agents | 58.2% | Not measured | Only GPT-5.6 Terra measured |
| METR time horizons | Not measured | 35.8% | Only Qwen2.5 72B Instruct measured |
| BALROG | 53.2% | 16.2% | GPT-5.6 Terra |
| TheAgentCompany | Not measured | 5.7% | Only Qwen2.5 72B Instruct measured |
| Vending-Bench 2 | $7,343 | Not measured | Only GPT-5.6 Terra measured |
| Neither model has a result on the other 18 boards here | |||
ReasoningLike-for-likePublic boards, shown as published, kept out of the index60
| Benchmark | GPT-5.6 Terra | Qwen2.5 72B Instruct | Leader |
|---|---|---|---|
| GPQA Diamond | 90.9% | Not measured | Only GPT-5.6 Terra measured |
| LiveBench reasoning | 90.6% | Not measured | Only GPT-5.6 Terra measured |
| MMMU Pro (Vals AI) | 86.5% | Not measured | Only GPT-5.6 Terra measured |
| GPQA Diamond (Artificial Analysis run) | 92.5% | 49.1% | GPT-5.6 Terra |
| Humanity's Last Exam (Artificial Analysis run) | 42.9% | 3.6% | GPT-5.6 Terra |
| GPQA Diamond (Epoch AI run) | 93.3% | 49.1% | GPT-5.6 Terra |
| ARC-AGI | 96.5% | Not measured | Only GPT-5.6 Terra measured |
| ARC-AGI-2 | 83.9% | Not measured | Only GPT-5.6 Terra measured |
| SimpleBench | 48.9% | Not measured | Only GPT-5.6 Terra measured |
| Chess puzzles | 47.0% | Not measured | Only GPT-5.6 Terra measured |
| Mystery game puzzles | 35.0% | Not measured | Only GPT-5.6 Terra measured |
| BIG-Bench Hard | Not measured | 79.8% | Only Qwen2.5 72B Instruct measured |
| LMArena Text: hard prompts | 1469 | 1271 | GPT-5.6 Terra |
| LMArena Text: expert | 1496 | 1245 | GPT-5.6 Terra |
| LMArena Text: hard prompts in English | 1465 | 1279 | GPT-5.6 Terra |
| LMArena Vision: diagrams | 1295 | Not measured | Only GPT-5.6 Terra measured |
| LMArena Vision: homework | 1312 | Not measured | Only GPT-5.6 Terra measured |
| Kagi LLM Benchmark | 51.3% | Not measured | Only GPT-5.6 Terra measured |
| Neither model has a result on the other 4 boards here | |||
MathLike-for-likePublic boards, shown as published, kept out of the index30
| Benchmark | GPT-5.6 Terra | Qwen2.5 72B Instruct | Leader |
|---|---|---|---|
| AA Math Index | Not measured | 14.0 | Only Qwen2.5 72B Instruct measured |
| AIME 2025 (Artificial Analysis run) | Not measured | 14.0% | Only Qwen2.5 72B Instruct measured |
| OTIS Mock AIME | 99.7% | 8.1% | GPT-5.6 Terra |
| FrontierMath, tiers 1 to 3 | 86.0% | Not measured | Only GPT-5.6 Terra measured |
| FrontierMath, tier 4 | 70.7% | Not measured | Only GPT-5.6 Terra measured |
| ProofBench | 74.0% | Not measured | Only GPT-5.6 Terra measured |
| MATH level 5 | Not measured | 63.2% | Only Qwen2.5 72B Instruct measured |
| LMArena Text: math | 1465 | 1283 | GPT-5.6 Terra |
| LMArena Text: mathematical work | 1478 | 1290 | GPT-5.6 Terra |
| Neither model has a result on the other 4 boards here | |||
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index40
| Benchmark | GPT-5.6 Terra | Qwen2.5 72B Instruct | Leader |
|---|---|---|---|
| MMLU Pro | 86.7% | Not measured | Only GPT-5.6 Terra measured |
| LegalBench (Vals AI) | 85.1% | Not measured | Only GPT-5.6 Terra measured |
| CorpFin (Vals AI) | 65.3% | Not measured | Only GPT-5.6 Terra measured |
| TaxEval (Vals AI) | 76.2% | Not measured | Only GPT-5.6 Terra measured |
| MortgageTax (Vals AI) | 67.3% | Not measured | Only GPT-5.6 Terra measured |
| MedCode (Vals AI) | 43.4% | Not measured | Only GPT-5.6 Terra measured |
| MMLU Pro (Artificial Analysis run) | Not measured | 72.0% | Only Qwen2.5 72B Instruct measured |
| SimpleQA Verified | 43.2% | Not measured | Only GPT-5.6 Terra measured |
| MMLU | Not measured | 85.3% | Only Qwen2.5 72B Instruct measured |
| AI2 Reasoning Challenge | Not measured | 94.5% | Only Qwen2.5 72B Instruct measured |
| LMArena Text: science | 1452 | 1275 | GPT-5.6 Terra |
| LMArena Text: medicine and healthcare | 1430 | 1258 | GPT-5.6 Terra |
| LMArena Text: legal and government | 1461 | 1288 | GPT-5.6 Terra |
| LMArena Text: business and finance | 1448 | 1253 | GPT-5.6 Terra |
| FACTS Benchmark Suite | 52.0% | Not measured | Only GPT-5.6 Terra measured |
| Neither model has a result on the other 3 boards here | |||
Web designDirectional onlyPublic boards, shown as published, kept out of the index9 boards, none shared
| Benchmark | GPT-5.6 Terra | Qwen2.5 72B Instruct | Leader |
|---|---|---|---|
| Design Arena: websites | 1268 | Not measured | Only GPT-5.6 Terra measured |
| Design Arena: UI components | 1256 | Not measured | Only GPT-5.6 Terra measured |
| Design Arena: game dev | 1229 | Not measured | Only GPT-5.6 Terra measured |
| Design Arena: data viz | 1240 | Not measured | Only GPT-5.6 Terra measured |
| Design Arena: 3D | 1134 | Not measured | Only GPT-5.6 Terra measured |
| Design Arena: all code categories | 1240 | Not measured | Only GPT-5.6 Terra measured |
| Design Arena: SVG | 1219 | Not measured | Only GPT-5.6 Terra measured |
| Design Arena: ASCII art | 1178 | Not measured | Only GPT-5.6 Terra measured |
| LiveBench design | 76.0% | Not measured | Only GPT-5.6 Terra measured |
GeneralLike-for-likePublic boards, shown as published, kept out of the index190
| Benchmark | GPT-5.6 Terra | Qwen2.5 72B Instruct | Leader |
|---|---|---|---|
| LiveBench data analysis | 79.3% | Not measured | Only GPT-5.6 Terra measured |
| Public Benefits Bench (Vals AI) | 62.4% | Not measured | Only GPT-5.6 Terra measured |
| SAGE (Vals AI) | 47.0% | Not measured | Only GPT-5.6 Terra measured |
| MedScribe (Vals AI) | 82.9% | Not measured | Only GPT-5.6 Terra measured |
| Vals Multimodal Index | 65.1% | Not measured | Only GPT-5.6 Terra measured |
| IFBench | 71.2% | 36.9% | GPT-5.6 Terra |
| Long context reasoning (AA-LCR) | 83.0% | Not measured | Only GPT-5.6 Terra measured |
| DTBench | 93.3% | 62.9% | GPT-5.6 Terra |
| LMCA | 54.9% | 13.4% | GPT-5.6 Terra |
| Surface Evolver Bench | 83.8% | Not measured | Only GPT-5.6 Terra measured |
| Furniture assembly | 54.2% | Not measured | Only GPT-5.6 Terra measured |
| HellaSwag | Not measured | 84.8% | Only Qwen2.5 72B Instruct measured |
| WinoGrande | Not measured | 82.3% | Only Qwen2.5 72B Instruct measured |
| PIQA | Not measured | 82.6% | Only Qwen2.5 72B Instruct measured |
| LMArena Text | 1448 | 1269 | GPT-5.6 Terra |
| LMArena Text: instruction following | 1456 | 1254 | GPT-5.6 Terra |
| LMArena Text: creative writing | 1413 | 1221 | GPT-5.6 Terra |
| LMArena Vision | 1271 | Not measured | Only GPT-5.6 Terra measured |
| LMArena Text: English | 1449 | 1283 | GPT-5.6 Terra |
| LMArena Text: non-English | 1441 | 1252 | GPT-5.6 Terra |
| LMArena Text: Chinese | 1514 | 1273 | GPT-5.6 Terra |
| LMArena Text: German | 1460 | 1234 | GPT-5.6 Terra |
| LMArena Text: Japanese | 1454 | 1180 | GPT-5.6 Terra |
| LMArena Text: Korean | 1427 | 1188 | GPT-5.6 Terra |
| LMArena Text: French | 1466 | 1280 | GPT-5.6 Terra |
| LMArena Text: Spanish | 1451 | 1255 | GPT-5.6 Terra |
| LMArena Text: Russian | 1452 | 1264 | GPT-5.6 Terra |
| LMArena Text: Polish | 1446 | Not measured | Only GPT-5.6 Terra measured |
| LMArena Text: multi-turn | 1453 | 1272 | GPT-5.6 Terra |
| LMArena Text: longer queries | 1454 | 1281 | GPT-5.6 Terra |
| LMArena Text: writing and language | 1431 | 1247 | GPT-5.6 Terra |
| LMArena Text: entertainment and media | 1418 | 1211 | GPT-5.6 Terra |
| LMArena Vision: English | 1272 | Not measured | Only GPT-5.6 Terra measured |
| LMArena Vision: Chinese | 1337 | Not measured | Only GPT-5.6 Terra measured |
| LMArena Vision: reading text | 1278 | Not measured | Only GPT-5.6 Terra measured |
| LMArena Vision: humor | 1236 | Not measured | Only GPT-5.6 Terra measured |
| LMArena Vision: creative writing | 1248 | Not measured | Only GPT-5.6 Terra measured |
| LMArena Document | 1472 | Not measured | Only GPT-5.6 Terra measured |
| Neither model has a result on the other 8 boards here | |||
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
| Benchmark | GPT-5.6 Terra | Qwen2.5 72B Instruct | Leader |
|---|---|---|---|
| Neither model has a result on 5 boards here | |||
What each workload costs
Three fixed token mixes at each model's listed price, so per-token rates become a bill.
Chat turn
1K input + 500 output tokens
GPT-5.6 Terra$0.0080Fits in one request
Qwen2.5 72B Instruct$0.0006Fits in one request
Qwen2.5 72B Instruct costs less
Repository review
50K input + 3K output tokens
GPT-5.6 Terra$0.136Fits in one request
Qwen2.5 72B Instruct$0.019Exceeds the context window
Qwen2.5 72B Instruct costs less
Agent loop
200K input + 10K output tokens
GPT-5.6 Terra$0.520Fits in one request
Qwen2.5 72B Instruct$0.076Exceeds the context window
Qwen2.5 72B Instruct costs less
Listed OpenRouter rates, without cache or batch discounts. A workload that does not fit a model's context window would need splitting.
Specification differences
The leading value in a row is marked; a row with a missing value names no leader. Our two rows are marked in red.
| Specification | OpenAI GPT-5.6 Terra | Qwen2.5 72B Instruct |
|---|---|---|
| Maker | OpenAI | Qwen |
| Weights | No public weights listed | Open weights |
| Released | July 9, 2026 | September 19, 2024 |
| Context window | 1.05M | 33K |
| Input $/Mtok | $2.00 | $0.36 |
| Output $/Mtok | $12.00 | $0.40 |
| Coding index | 63.9 | Not benchmarked |
| Independent measurements behind it | 6 | 0 |
| Cost per solved task | $0.42 | Not computable |
| Source | OpenRouter listing | OpenRouter listing |
Questions
Is GPT-5.6 Terra better than Qwen2.5 72B Instruct for coding?
Only GPT-5.6 Terra has a coding score, 63.9 on the Vibe Coding Index; Qwen2.5 72B Instruct has none, so the two cannot be ranked on coding. Pick Qwen2.5 72B Instruct for output tokens 30.0x cheaper ($0.40 vs $12.00 per million).
Which is cheaper, GPT-5.6 Terra or Qwen2.5 72B Instruct?
GPT-5.6 Terra costs $2.00 per million input tokens and $12.00 per million output tokens; Qwen2.5 72B Instruct costs $0.36 and $0.40. Cost per solved task is known for GPT-5.6 Terra only.
Which has the bigger context window, GPT-5.6 Terra or Qwen2.5 72B Instruct?
GPT-5.6 Terra, with 1.05M tokens against 33K.
How many benchmarks is the GPT-5.6 Terra vs Qwen2.5 72B Instruct comparison based on?
40 benchmarks have a result for both models as of October 22, 2026. GPT-5.6 Terra's coding index rests on 6 independent benchmarks and Qwen2.5 72B Instruct's on 0. Where only one model has a result, the page says so instead of naming a winner.