GPT-5.6 Terra vs Qwen3 235B A22B Instruct 2507
GPT-5.6 Terra leads Qwen3 235B A22B Instruct 2507 on the Artificial Analysis Intelligence Index, 42.1 to 12.0, and Qwen3 235B A22B Instruct 2507 costs 15.5x less per output token ($0.775 vs $12.00 per million). Updated October 23, 2026; 19 benchmarks have a result for both.
vs boards won of 19 shared
GPT-5.6 Terra or Qwen3 235B A22B Instruct 2507: which to pick
Only GPT-5.6 Terra has a coding score, 63.9 on the Vibe Coding Index; Qwen3 235B A22B Instruct 2507 has none, so the two cannot be ranked on coding.
Pick Qwen3 235B A22B Instruct 2507 for output tokens 15.5x cheaper ($0.775 vs $12.00 per million).
Model AGPT-5.6 TerraOpenAI63.9/1006 independent benchmarks$0.42 per solved taskModel BQwen3 235B A22B Instruct 2507QwenNot benchmarkedNo coding benchmarkNo cost per solved task
- Shared results
- 19
- GPT-5.6 Terra only
- 110
- Qwen3 235B A22B Instruct 2507 only
- 5
- Like-for-like categories
- 6 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 12.0.
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
GPT-5.6 Terra
GPT-5.6 Terra leads, 1268 against 1058.
Same board, both measured
Long documents
Listed context window
GPT-5.6 Terra
GPT-5.6 Terra takes 1.05M tokens against 262K.
Listed by OpenRouter
Everyday cost
1K input + 500 output tokens
Qwen3 235B A22B Instruct 2507
GPT-5.6 Terra $0.0080, Qwen3 235B A22B Instruct 2507 $0.0005 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, Qwen3 235B A22B Instruct 2507 not benchmarked.
- Widest category
- Reasoning, 52.2 points apart on average over 4 shared benchmarks.
- Widest single gap
- ARC-AGI: GPT-5.6 Terra 96.5% against Qwen3 235B A22B Instruct 2507 11.0%, 85.5 points apart on a 0 to 100 scale.
- Output price
- GPT-5.6 Terra $12.00 against Qwen3 235B A22B Instruct 2507 $0.775 per million output tokens, 15.5x apart.
- Context window
- GPT-5.6 Terra takes 1.05M tokens, Qwen3 235B A22B Instruct 2507 262K.
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 | Qwen3 235B A22B Instruct 2507 | Ties |
|---|---|---|---|---|
| Coding | 3 | 0 | ||
| Intelligence | 1 | 0 | ||
| Agentic work | 1 | 0 | ||
| Reasoning | 4 | 0 | ||
| Web design | 6 | 0 | ||
| General | 4 | 0 | ||
| All shared boards | 19 | 0 |
CodingLike-for-likeSix of these boards feed the Vibe Coding Index30
| Benchmark | GPT-5.6 Terra | Qwen3 235B A22B Instruct 2507 | Leader |
|---|---|---|---|
| LiveBench coding | 86.2% | Not measured | Only GPT-5.6 Terra measured |
| LMArena WebDev | 1521 | 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 | 78.8% | Only Qwen3 235B A22B Instruct 2507 measured |
| SciCode | 55.0% | 41.4% | GPT-5.6 Terra |
| Terminal-Bench Hard (Artificial Analysis run) | 62.9% | 15.2% | GPT-5.6 Terra |
| FrontierCode | 41.3% | Not measured | Only GPT-5.6 Terra measured |
| WeirdML | 78.3% | 38.7% | GPT-5.6 Terra |
| ALE-Bench | 1951.4 | Not measured | Only GPT-5.6 Terra measured |
| LMArena Text: coding | 1486 | Not measured | Only GPT-5.6 Terra measured |
| LMArena Text: software and IT | 1482 | Not measured | Only GPT-5.6 Terra measured |
| LMArena WebDev: React | 1511 | 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 |
| SWE-Bench Pro V2 | 92.4% | Not measured | Only GPT-5.6 Terra measured |
| Neither model has a result on the other 16 boards here | |||
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index10
| Benchmark | GPT-5.6 Terra | Qwen3 235B A22B Instruct 2507 | Leader |
|---|---|---|---|
| AA Intelligence Index | 42.1 | 12.0 | GPT-5.6 Terra |
| Epoch Capabilities Index | 160.1 | Not measured | Only GPT-5.6 Terra measured |
Agentic workLike-for-likePublic boards, shown as published, kept out of the index10
| Benchmark | GPT-5.6 Terra | Qwen3 235B A22B Instruct 2507 | 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 | 1110 | 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 | 1081 | Not measured | Only GPT-5.6 Terra measured |
| Design Arena: HTML slides | 1013 | Not measured | Only GPT-5.6 Terra measured |
| Design Arena: PowerPoint slides | 1095 | 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% | 53.2% | GPT-5.6 Terra |
| APEX-Agents | 58.2% | Not measured | Only GPT-5.6 Terra measured |
| BALROG | 53.2% | Not measured | Only GPT-5.6 Terra measured |
| Vending-Bench 2 | $7,343 | Not measured | Only GPT-5.6 Terra measured |
| Berkeley Function Calling Leaderboard (BFCL) | Not measured | 52.1% | Only Qwen3 235B A22B Instruct 2507 measured |
| Neither model has a result on the other 20 boards here | |||
ReasoningLike-for-likePublic boards, shown as published, kept out of the index40
| Benchmark | GPT-5.6 Terra | Qwen3 235B A22B Instruct 2507 | 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% | 79.0% | GPT-5.6 Terra |
| Humanity's Last Exam (Artificial Analysis run) | 42.9% | 15.9% | GPT-5.6 Terra |
| GPQA Diamond (Epoch AI run) | 93.3% | Not measured | Only GPT-5.6 Terra measured |
| ARC-AGI | 96.5% | 11.0% | GPT-5.6 Terra |
| ARC-AGI-2 | 83.9% | 1.3% | GPT-5.6 Terra |
| 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 |
| LMArena Text: hard prompts | 1469 | Not measured | Only GPT-5.6 Terra measured |
| LMArena Text: expert | 1496 | Not measured | Only GPT-5.6 Terra measured |
| LMArena Text: hard prompts in English | 1465 | Not measured | Only GPT-5.6 Terra measured |
| 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 6 boards here | |||
MathDirectional onlyPublic boards, shown as published, kept out of the index8 boards, none shared
| Benchmark | GPT-5.6 Terra | Qwen3 235B A22B Instruct 2507 | Leader |
|---|---|---|---|
| AA Math Index | Not measured | 91.0 | Only Qwen3 235B A22B Instruct 2507 measured |
| AIME 2025 (Artificial Analysis run) | Not measured | 91.0% | Only Qwen3 235B A22B Instruct 2507 measured |
| OTIS Mock AIME | 99.7% | Not measured | Only GPT-5.6 Terra measured |
| 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 |
| LMArena Text: math | 1466 | Not measured | Only GPT-5.6 Terra measured |
| LMArena Text: mathematical work | 1478 | Not measured | Only GPT-5.6 Terra measured |
| Neither model has a result on the other 6 boards here | |||
KnowledgeDirectional onlyPublic boards, shown as published, kept out of the index13 boards, none shared
| Benchmark | GPT-5.6 Terra | Qwen3 235B A22B Instruct 2507 | 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 | 84.3% | Only Qwen3 235B A22B Instruct 2507 measured |
| SimpleQA Verified | 43.2% | Not measured | Only GPT-5.6 Terra measured |
| LMArena Text: science | 1452 | Not measured | Only GPT-5.6 Terra measured |
| LMArena Text: medicine and healthcare | 1432 | Not measured | Only GPT-5.6 Terra measured |
| LMArena Text: legal and government | 1462 | Not measured | Only GPT-5.6 Terra measured |
| LMArena Text: business and finance | 1448 | Not measured | Only GPT-5.6 Terra measured |
| FACTS Benchmark Suite | 52.0% | Not measured | Only GPT-5.6 Terra measured |
| Neither model has a result on the other 7 boards here | |||
Web designLike-for-likePublic boards, shown as published, kept out of the index60
| Benchmark | GPT-5.6 Terra | Qwen3 235B A22B Instruct 2507 | Leader |
|---|---|---|---|
| Design Arena: websites | 1268 | 1058 | GPT-5.6 Terra |
| Design Arena: UI components | 1256 | 963 | GPT-5.6 Terra |
| Design Arena: game dev | 1229 | 951 | GPT-5.6 Terra |
| Design Arena: data viz | 1240 | 1070 | GPT-5.6 Terra |
| Design Arena: 3D | 1135 | 995 | GPT-5.6 Terra |
| Design Arena: all code categories | 1241 | 1038 | GPT-5.6 Terra |
| 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 index40
| Benchmark | GPT-5.6 Terra | Qwen3 235B A22B Instruct 2507 | 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% | 51.2% | GPT-5.6 Terra |
| Long context reasoning (AA-LCR) | 83.0% | 72.0% | GPT-5.6 Terra |
| DTBench | 93.3% | 78.4% | GPT-5.6 Terra |
| LMCA | 54.9% | 23.5% | 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 |
| LMArena Text | 1448 | Not measured | Only GPT-5.6 Terra measured |
| LMArena Text: instruction following | 1455 | Not measured | Only GPT-5.6 Terra measured |
| LMArena Text: creative writing | 1413 | Not measured | Only GPT-5.6 Terra measured |
| LMArena Vision | 1271 | Not measured | Only GPT-5.6 Terra measured |
| LMArena Text: English | 1449 | Not measured | Only GPT-5.6 Terra measured |
| LMArena Text: non-English | 1441 | Not measured | Only GPT-5.6 Terra measured |
| LMArena Text: Chinese | 1514 | Not measured | Only GPT-5.6 Terra measured |
| LMArena Text: German | 1461 | Not measured | Only GPT-5.6 Terra measured |
| LMArena Text: Japanese | 1453 | Not measured | Only GPT-5.6 Terra measured |
| LMArena Text: Korean | 1423 | Not measured | Only GPT-5.6 Terra measured |
| LMArena Text: French | 1466 | Not measured | Only GPT-5.6 Terra measured |
| LMArena Text: Spanish | 1448 | Not measured | Only GPT-5.6 Terra measured |
| LMArena Text: Russian | 1453 | Not measured | Only GPT-5.6 Terra measured |
| LMArena Text: Polish | 1447 | Not measured | Only GPT-5.6 Terra measured |
| LMArena Text: multi-turn | 1453 | Not measured | Only GPT-5.6 Terra measured |
| LMArena Text: longer queries | 1453 | Not measured | Only GPT-5.6 Terra measured |
| LMArena Text: writing and language | 1432 | Not measured | Only GPT-5.6 Terra measured |
| LMArena Text: entertainment and media | 1419 | Not measured | Only GPT-5.6 Terra measured |
| 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 |
| EQ-Bench 4 | 1234 | Not measured | Only GPT-5.6 Terra measured |
| Neither model has a result on the other 12 boards here | |||
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
| Benchmark | GPT-5.6 Terra | Qwen3 235B A22B Instruct 2507 | 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
Qwen3 235B A22B Instruct 2507$0.0005Fits in one request
Qwen3 235B A22B Instruct 2507 costs less
Repository review
50K input + 3K output tokens
GPT-5.6 Terra$0.136Fits in one request
Qwen3 235B A22B Instruct 2507$0.0096Fits in one request
Qwen3 235B A22B Instruct 2507 costs less
Agent loop
200K input + 10K output tokens
GPT-5.6 Terra$0.520Fits in one request
Qwen3 235B A22B Instruct 2507$0.037Fits in one request
Qwen3 235B A22B Instruct 2507 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 | Qwen3 235B A22B Instruct 2507 |
|---|---|---|
| Maker | OpenAI | Qwen |
| Weights | No public weights listed | Open weights |
| Released | July 9, 2026 | July 21, 2025 |
| Context window | 1.05M | 262K |
| Input $/Mtok | $2.00 | $0.145 |
| Output $/Mtok | $12.00 | $0.775 |
| 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 Qwen3 235B A22B Instruct 2507 for coding?
Only GPT-5.6 Terra has a coding score, 63.9 on the Vibe Coding Index; Qwen3 235B A22B Instruct 2507 has none, so the two cannot be ranked on coding. Pick Qwen3 235B A22B Instruct 2507 for output tokens 15.5x cheaper ($0.775 vs $12.00 per million).
Which is cheaper, GPT-5.6 Terra or Qwen3 235B A22B Instruct 2507?
GPT-5.6 Terra costs $2.00 per million input tokens and $12.00 per million output tokens; Qwen3 235B A22B Instruct 2507 costs $0.145 and $0.775. Cost per solved task is known for GPT-5.6 Terra only.
Which has the bigger context window, GPT-5.6 Terra or Qwen3 235B A22B Instruct 2507?
GPT-5.6 Terra, with 1.05M tokens against 262K.
How many benchmarks is the GPT-5.6 Terra vs Qwen3 235B A22B Instruct 2507 comparison based on?
19 benchmarks have a result for both models as of October 23, 2026. GPT-5.6 Terra's coding index rests on 6 independent benchmarks and Qwen3 235B A22B Instruct 2507's on 0. Where only one model has a result, the page says so instead of naming a winner.