UI-TARS 7B vs Claude Sonnet 5
UI-TARS 7B costs 50.0x less per output token than Claude Sonnet 5 ($0.20 vs $10.00 per million). Updated October 22, 2026; no benchmark has tested both models yet.
UI-TARS 7BByteDance
- Coding
- No data
- Overall
- No data
- Output $/M
- $0.20
Claude Sonnet 5Anthropic
- Coding
- 57.7
- Overall
- 63.2
- Output $/M
- $10.00
vs $0.20 vs $10.00 per 1M output tokens. UI-TARS 7B is 50.0x cheaper.
UI-TARS 7B or Claude Sonnet 5: which to pick
Only Claude Sonnet 5 has a coding score, 57.7 on the Vibe Coding Index; UI-TARS 7B has none, so the two cannot be ranked on coding.
Pick UI-TARS 7B for output tokens 50.0x cheaper ($0.20 vs $10.00 per million).
Model AUI-TARS 7BByteDanceNot benchmarkedNo coding benchmarkNo cost per solved taskModel BClaude Sonnet 5Anthropic57.7/1006 independent benchmarks$1.87 per solved task
No benchmark has tested both yet. UI-TARS 7B has results on Kagi LLM Benchmark; Claude Sonnet 5 on LiveBench coding, LMArena WebDev and AA Coding Index. Compare them on price and context instead.
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 Claude Sonnet 5 has a coding score, so no call is made.
One side not measured
Intelligence
AA Intelligence Index
No call
Only Claude Sonnet 5 has a AA Intelligence Index result, 38.2.
Directional only
Agentic work
AA Agentic Index
No call
Only Claude Sonnet 5 has a AA Agentic Index result, 43.6.
Directional only
Reasoning
GPQA Diamond
No call
Only Claude Sonnet 5 has a GPQA Diamond result, 88.9%.
Directional only
Knowledge
MMLU Pro
No call
Only Claude Sonnet 5 has a MMLU Pro result, 87.5%.
Directional only
Web design
Design Arena: websites
No call
Only Claude Sonnet 5 has a Design Arena: websites result, 1274.
Directional only
Long documents
Listed context window
Claude Sonnet 5
Claude Sonnet 5 takes 1M tokens against 128K.
Listed by OpenRouter
Everyday cost
1K input + 500 output tokens
UI-TARS 7B
UI-TARS 7B $0.0002, Claude Sonnet 5 $0.0070 per turn.
List prices
Cost per solved task
One SWE-bench Verified task, solved
No call
Only Claude Sonnet 5 has the number, $1.87.
Directional only
At a glance
- Coding index
- UI-TARS 7B not benchmarked, Claude Sonnet 5 57.7 from 6 benchmarks.
- Widest category
- None yet: no category has been tested on both models.
- Widest single gap
- None yet: no benchmark has tested both models.
- Output price
- UI-TARS 7B $0.20 against Claude Sonnet 5 $10.00 per million output tokens, 50.0x apart.
- Context window
- UI-TARS 7B takes 128K tokens, Claude Sonnet 5 1M.
Benchmark by benchmark
Every board either model has a result on, a category to a fold; bars are each result on a 0 to 100 scale. Only one model is measured on each, so no row names a leader.
CodingDirectional onlySix of these boards feed the Vibe Coding Index25 boards measured
| Benchmark | UI-TARS 7B | Claude Sonnet 5 | Leader |
|---|---|---|---|
| LiveBench coding | Not measured | 74.7% | Only Claude Sonnet 5 measured |
| LMArena WebDev | Not measured | 1540 | Only Claude Sonnet 5 measured |
| AA Coding Index | Not measured | 71.5 | Only Claude Sonnet 5 measured |
| LiveCodeBench (Vals AI) | Not measured | 82.4% | Only Claude Sonnet 5 measured |
| SWE-bench Verified (Vals AI) | Not measured | 79.6% | Only Claude Sonnet 5 measured |
| Terminal-Bench 4 (Vals AI) | Not measured | 10.6% | Only Claude Sonnet 5 measured |
| Vibe Code Bench (Vals AI) | Not measured | 81.3% | Only Claude Sonnet 5 measured |
| Code migration (Vals AI) | Not measured | 44.4% | Only Claude Sonnet 5 measured |
| ProgramBench (Vals AI) | Not measured | 0.0% | Only Claude Sonnet 5 measured |
| IOI (Vals AI) | Not measured | 45.0% | Only Claude Sonnet 5 measured |
| Terminal-Bench 2.1 (Vals AI) | Not measured | 74.5% | Only Claude Sonnet 5 measured |
| Vibe Code Bench 1-100 (Vals AI) | Not measured | 13.8% | Only Claude Sonnet 5 measured |
| SkillsBench (Vals AI) | Not measured | 46.5% | Only Claude Sonnet 5 measured |
| CyberBench (Vals AI) | Not measured | 70.3% | Only Claude Sonnet 5 measured |
| SciCode | Not measured | 54.3% | Only Claude Sonnet 5 measured |
| FrontierCode | Not measured | 42.7% | Only Claude Sonnet 5 measured |
| GSO-Bench | Not measured | 37.3% | Only Claude Sonnet 5 measured |
| WeirdML | Not measured | 68.8% | Only Claude Sonnet 5 measured |
| ALE-Bench | Not measured | 1463.1 | Only Claude Sonnet 5 measured |
| LMArena Text: coding | Not measured | 1483 | Only Claude Sonnet 5 measured |
| LMArena Text: software and IT | Not measured | 1474 | Only Claude Sonnet 5 measured |
| LMArena WebDev: React | Not measured | 1541 | Only Claude Sonnet 5 measured |
| LMArena WebDev: HTML | Not measured | 1527 | Only Claude Sonnet 5 measured |
| LMArena WebDev: from an image | Not measured | 1540 | Only Claude Sonnet 5 measured |
| SciCode | Not measured | 10.7% | Only Claude Sonnet 5 measured |
| Neither model has a result on the other 14 boards here | |||
IntelligenceDirectional onlyPublic boards, shown as published, kept out of the index2 boards measured
| Benchmark | UI-TARS 7B | Claude Sonnet 5 | Leader |
|---|---|---|---|
| AA Intelligence Index | Not measured | 38.2 | Only Claude Sonnet 5 measured |
| Epoch Capabilities Index | Not measured | 156.9 | Only Claude Sonnet 5 measured |
Agentic workDirectional onlyPublic boards, shown as published, kept out of the index21 boards measured
| Benchmark | UI-TARS 7B | Claude Sonnet 5 | Leader |
|---|---|---|---|
| AA Agentic Index | Not measured | 43.6 | Only Claude Sonnet 5 measured |
| Design Arena: agents building web apps | Not measured | 1245 | Only Claude Sonnet 5 measured |
| Design Arena: agents building full-stack apps | Not measured | 1249 | Only Claude Sonnet 5 measured |
| Design Arena: agents building mobile apps | Not measured | 1214 | Only Claude Sonnet 5 measured |
| Design Arena: agents building Android apps | Not measured | 1227 | Only Claude Sonnet 5 measured |
| Design Arena: agents building Godot games | Not measured | 1225 | Only Claude Sonnet 5 measured |
| Design Arena: agents building games | Not measured | 1227 | Only Claude Sonnet 5 measured |
| Design Arena: HTML slides | Not measured | 1168 | Only Claude Sonnet 5 measured |
| Design Arena: PowerPoint slides | Not measured | 1193 | Only Claude Sonnet 5 measured |
| Design Arena: spreadsheets | Not measured | 1077 | Only Claude Sonnet 5 measured |
| LiveBench research | Not measured | 82.8% | Only Claude Sonnet 5 measured |
| LiveBench browser | Not measured | 72.0% | Only Claude Sonnet 5 measured |
| LiveBench long agentic | Not measured | 39.0% | Only Claude Sonnet 5 measured |
| SREBench (Vals AI) | Not measured | 0.4% | Only Claude Sonnet 5 measured |
| Finance Agent (Vals AI) | Not measured | 53.9% | Only Claude Sonnet 5 measured |
| Tax Agent Bench (Vals AI) | Not measured | 29.4% | Only Claude Sonnet 5 measured |
| Legal Research Bench (Vals AI) | Not measured | 41.8% | Only Claude Sonnet 5 measured |
| Excel Modeling Benchmark (Vals AI) | Not measured | 66.3% | Only Claude Sonnet 5 measured |
| Terminal-Bench-Science (Vals AI) | Not measured | 5.7% | Only Claude Sonnet 5 measured |
| APEX-Agents | Not measured | 54.5% | Only Claude Sonnet 5 measured |
| Vending-Bench 2 | Not measured | $6,378 | Only Claude Sonnet 5 measured |
| Neither model has a result on the other 20 boards here | |||
ReasoningDirectional onlyPublic boards, shown as published, kept out of the index15 boards measured
| Benchmark | UI-TARS 7B | Claude Sonnet 5 | Leader |
|---|---|---|---|
| GPQA Diamond | Not measured | 88.9% | Only Claude Sonnet 5 measured |
| LiveBench reasoning | Not measured | 88.7% | Only Claude Sonnet 5 measured |
| MMMU Pro (Vals AI) | Not measured | 83.0% | Only Claude Sonnet 5 measured |
| GPQA Diamond (Artificial Analysis run) | Not measured | 91.1% | Only Claude Sonnet 5 measured |
| Humanity's Last Exam (Artificial Analysis run) | Not measured | 41.3% | Only Claude Sonnet 5 measured |
| GPQA Diamond (Epoch AI run) | Not measured | 90.5% | Only Claude Sonnet 5 measured |
| SimpleBench | Not measured | 60.6% | Only Claude Sonnet 5 measured |
| Chess puzzles | Not measured | 45.0% | Only Claude Sonnet 5 measured |
| Mystery game puzzles | Not measured | 35.0% | Only Claude Sonnet 5 measured |
| LMArena Text: hard prompts | Not measured | 1461 | Only Claude Sonnet 5 measured |
| LMArena Text: expert | Not measured | 1489 | Only Claude Sonnet 5 measured |
| LMArena Text: hard prompts in English | Not measured | 1464 | Only Claude Sonnet 5 measured |
| LMArena Vision: diagrams | Not measured | 1293 | Only Claude Sonnet 5 measured |
| LMArena Vision: homework | Not measured | 1311 | Only Claude Sonnet 5 measured |
| Kagi LLM Benchmark | 20.7% | Not measured | Only UI-TARS 7B measured |
| Neither model has a result on the other 7 boards here | |||
MathDirectional onlyPublic boards, shown as published, kept out of the index6 boards measured
| Benchmark | UI-TARS 7B | Claude Sonnet 5 | Leader |
|---|---|---|---|
| OTIS Mock AIME | Not measured | 94.7% | Only Claude Sonnet 5 measured |
| FrontierMath, tiers 1 to 3 | Not measured | 65.6% | Only Claude Sonnet 5 measured |
| FrontierMath, tier 4 | Not measured | 29.3% | Only Claude Sonnet 5 measured |
| ProofBench | Not measured | 77.0% | Only Claude Sonnet 5 measured |
| LMArena Text: math | Not measured | 1467 | Only Claude Sonnet 5 measured |
| LMArena Text: mathematical work | Not measured | 1481 | Only Claude Sonnet 5 measured |
| Neither model has a result on the other 7 boards here | |||
KnowledgeDirectional onlyPublic boards, shown as published, kept out of the index11 boards measured
| Benchmark | UI-TARS 7B | Claude Sonnet 5 | Leader |
|---|---|---|---|
| MMLU Pro | Not measured | 87.5% | Only Claude Sonnet 5 measured |
| LegalBench (Vals AI) | Not measured | 83.9% | Only Claude Sonnet 5 measured |
| CorpFin (Vals AI) | Not measured | 67.0% | Only Claude Sonnet 5 measured |
| TaxEval (Vals AI) | Not measured | 75.6% | Only Claude Sonnet 5 measured |
| MortgageTax (Vals AI) | Not measured | 70.0% | Only Claude Sonnet 5 measured |
| MedCode (Vals AI) | Not measured | 47.5% | Only Claude Sonnet 5 measured |
| SimpleQA Verified | Not measured | 33.7% | Only Claude Sonnet 5 measured |
| LMArena Text: science | Not measured | 1460 | Only Claude Sonnet 5 measured |
| LMArena Text: medicine and healthcare | Not measured | 1442 | Only Claude Sonnet 5 measured |
| LMArena Text: legal and government | Not measured | 1454 | Only Claude Sonnet 5 measured |
| LMArena Text: business and finance | Not measured | 1436 | Only Claude Sonnet 5 measured |
| Neither model has a result on the other 7 boards here | |||
Web designDirectional onlyPublic boards, shown as published, kept out of the index9 boards measured
| Benchmark | UI-TARS 7B | Claude Sonnet 5 | Leader |
|---|---|---|---|
| Design Arena: websites | Not measured | 1274 | Only Claude Sonnet 5 measured |
| Design Arena: UI components | Not measured | 1277 | Only Claude Sonnet 5 measured |
| Design Arena: game dev | Not measured | 1276 | Only Claude Sonnet 5 measured |
| Design Arena: data viz | Not measured | 1252 | Only Claude Sonnet 5 measured |
| Design Arena: 3D | Not measured | 1260 | Only Claude Sonnet 5 measured |
| Design Arena: all code categories | Not measured | 1275 | Only Claude Sonnet 5 measured |
| Design Arena: SVG | Not measured | 1192 | Only Claude Sonnet 5 measured |
| Design Arena: ASCII art | Not measured | 1207 | Only Claude Sonnet 5 measured |
| LiveBench design | Not measured | 82.0% | Only Claude Sonnet 5 measured |
GeneralDirectional onlyPublic boards, shown as published, kept out of the index33 boards measured
| Benchmark | UI-TARS 7B | Claude Sonnet 5 | Leader |
|---|---|---|---|
| LiveBench data analysis | Not measured | 71.7% | Only Claude Sonnet 5 measured |
| Public Benefits Bench (Vals AI) | Not measured | 66.0% | Only Claude Sonnet 5 measured |
| SAGE (Vals AI) | Not measured | 48.9% | Only Claude Sonnet 5 measured |
| MedScribe (Vals AI) | Not measured | 76.0% | Only Claude Sonnet 5 measured |
| Vals Multimodal Index | Not measured | 68.8% | Only Claude Sonnet 5 measured |
| Long context reasoning (AA-LCR) | Not measured | 82.0% | Only Claude Sonnet 5 measured |
| DTBench | Not measured | 92.5% | Only Claude Sonnet 5 measured |
| LMCA | Not measured | 50.0% | Only Claude Sonnet 5 measured |
| Surface Evolver Bench | Not measured | 60.0% | Only Claude Sonnet 5 measured |
| LMArena Text | Not measured | 1442 | Only Claude Sonnet 5 measured |
| LMArena Text: instruction following | Not measured | 1452 | Only Claude Sonnet 5 measured |
| LMArena Text: creative writing | Not measured | 1416 | Only Claude Sonnet 5 measured |
| LMArena Vision | Not measured | 1273 | Only Claude Sonnet 5 measured |
| LMArena Text: English | Not measured | 1450 | Only Claude Sonnet 5 measured |
| LMArena Text: non-English | Not measured | 1431 | Only Claude Sonnet 5 measured |
| LMArena Text: Chinese | Not measured | 1478 | Only Claude Sonnet 5 measured |
| LMArena Text: German | Not measured | 1442 | Only Claude Sonnet 5 measured |
| LMArena Text: Japanese | Not measured | 1422 | Only Claude Sonnet 5 measured |
| LMArena Text: Korean | Not measured | 1412 | Only Claude Sonnet 5 measured |
| LMArena Text: French | Not measured | 1463 | Only Claude Sonnet 5 measured |
| LMArena Text: Spanish | Not measured | 1438 | Only Claude Sonnet 5 measured |
| LMArena Text: Russian | Not measured | 1450 | Only Claude Sonnet 5 measured |
| LMArena Text: Polish | Not measured | 1425 | Only Claude Sonnet 5 measured |
| LMArena Text: multi-turn | Not measured | 1453 | Only Claude Sonnet 5 measured |
| LMArena Text: longer queries | Not measured | 1462 | Only Claude Sonnet 5 measured |
| LMArena Text: writing and language | Not measured | 1436 | Only Claude Sonnet 5 measured |
| LMArena Text: entertainment and media | Not measured | 1411 | Only Claude Sonnet 5 measured |
| LMArena Vision: English | Not measured | 1275 | Only Claude Sonnet 5 measured |
| LMArena Vision: Chinese | Not measured | 1331 | Only Claude Sonnet 5 measured |
| LMArena Vision: reading text | Not measured | 1288 | Only Claude Sonnet 5 measured |
| LMArena Vision: humor | Not measured | 1261 | Only Claude Sonnet 5 measured |
| LMArena Vision: creative writing | Not measured | 1246 | Only Claude Sonnet 5 measured |
| LMArena Document | Not measured | 1466 | Only Claude Sonnet 5 measured |
| Neither model has a result on the other 13 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
UI-TARS 7B$0.0002Fits in one request
Claude Sonnet 5$0.0070Fits in one request
UI-TARS 7B costs less
Repository review
50K input + 3K output tokens
UI-TARS 7B$0.0056Fits in one request
Claude Sonnet 5$0.130Fits in one request
UI-TARS 7B costs less
Agent loop
200K input + 10K output tokens
UI-TARS 7B$0.022Exceeds the context window
Claude Sonnet 5$0.500Fits in one request
UI-TARS 7B 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 | ByteDance UI-TARS 7B | Anthropic Claude Sonnet 5 |
|---|---|---|
| Maker | ByteDance | Anthropic |
| Weights | Open weights | No public weights listed |
| Released | July 22, 2025 | June 30, 2026 |
| Context window | 128K | 1M |
| Input $/Mtok | $0.10 | $2.00 |
| Output $/Mtok | $0.20 | $10.00 |
| Coding index | Not benchmarked | 57.7 |
| Independent measurements behind it | 0 | 6 |
| Cost per solved task | Not computable | $1.87 |
| Source | OpenRouter listing | OpenRouter listing |
Questions
Is UI-TARS 7B better than Claude Sonnet 5 for coding?
Only Claude Sonnet 5 has a coding score, 57.7 on the Vibe Coding Index; UI-TARS 7B has none, so the two cannot be ranked on coding. Pick UI-TARS 7B for output tokens 50.0x cheaper ($0.20 vs $10.00 per million).
Which is cheaper, UI-TARS 7B or Claude Sonnet 5?
UI-TARS 7B costs $0.10 per million input tokens and $0.20 per million output tokens; Claude Sonnet 5 costs $2.00 and $10.00. Cost per solved task is known for Claude Sonnet 5 only.
Which has the bigger context window, UI-TARS 7B or Claude Sonnet 5?
Claude Sonnet 5, with 1M tokens against 128K.
Has any benchmark tested both UI-TARS 7B and Claude Sonnet 5?
No benchmark has tested both yet. UI-TARS 7B has results on Kagi LLM Benchmark; Claude Sonnet 5 on LiveBench coding, LMArena WebDev and AA Coding Index. Compare them on price and context instead. Data as of October 22, 2026.