UI-TARS 7B vs Claude Opus 4.1
UI-TARS 7B costs 375.0x less per output token than Claude Opus 4.1 ($0.20 vs $75.00 per million). Updated October 22, 2026; no benchmark has tested both models yet.
- Coding
- No data
- Overall
- No data
- Output $/M
- $0.20
- Coding
- No data
- Overall
- 50.2
- Output $/M
- $75.00
UI-TARS 7B or Claude Opus 4.1: which to pick
Neither model has a coding score yet, so the two cannot be ranked on coding.
Pick UI-TARS 7B for output tokens 375.0x cheaper ($0.20 vs $75.00 per million).
No benchmark has tested both yet. UI-TARS 7B has results on Kagi LLM Benchmark; Claude Opus 4.1 on LMArena WebDev, LiveCodeBench (Vals AI) and Terminal-Bench. 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
Neither has a coding score yet.
One side not measured
Intelligence
AA Intelligence Index
No call
Only Claude Opus 4.1 has a AA Intelligence Index result, 22.8.
Directional only
Reasoning
GPQA Diamond
No call
Only Claude Opus 4.1 has a GPQA Diamond result, 76.3%.
Directional only
Math
AIME
No call
Only Claude Opus 4.1 has a AIME result, 78.2%.
Directional only
Knowledge
MMLU Pro
No call
Only Claude Opus 4.1 has a MMLU Pro result, 87.9%.
Directional only
Web design
Design Arena: websites
No call
Only Claude Opus 4.1 has a Design Arena: websites result, 1176.
Directional only
Long documents
Listed context window
Claude Opus 4.1
Claude Opus 4.1 takes 200K tokens against 128K.
Listed by OpenRouter
Everyday cost
1K input + 500 output tokens
UI-TARS 7B
UI-TARS 7B $0.0002, Claude Opus 4.1 $0.052 per turn.
List prices
At a glance
- Coding index
- UI-TARS 7B not benchmarked, Claude Opus 4.1 not benchmarked.
- 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 Opus 4.1 $75.00 per million output tokens, 375.0x apart.
- Context window
- UI-TARS 7B takes 128K tokens, Claude Opus 4.1 200K.
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 Index9 boards measured
| Benchmark | UI-TARS 7B | Claude Opus 4.1 | Leader |
|---|---|---|---|
| LMArena WebDev | Not measured | 1390 | Only Claude Opus 4.1 measured |
| LiveCodeBench (Vals AI) | Not measured | 66.5% | Only Claude Opus 4.1 measured |
| Terminal-Bench | Not measured | 38.0% | Only Claude Opus 4.1 measured |
| WeirdML | Not measured | 45.9% | Only Claude Opus 4.1 measured |
| Cybench | Not measured | 42.0% | Only Claude Opus 4.1 measured |
| LMArena Text: coding | Not measured | 1479 | Only Claude Opus 4.1 measured |
| LMArena Text: software and IT | Not measured | 1456 | Only Claude Opus 4.1 measured |
| LMArena WebDev: HTML | Not measured | 1399 | Only Claude Opus 4.1 measured |
| SciCode | Not measured | 4.6% | Only Claude Opus 4.1 measured |
| Neither model has a result on the other 34 boards here | |||
IntelligenceDirectional onlyPublic boards, shown as published, kept out of the index2 boards measured
| Benchmark | UI-TARS 7B | Claude Opus 4.1 | Leader |
|---|---|---|---|
| AA Intelligence Index | Not measured | 22.8 | Only Claude Opus 4.1 measured |
| Epoch Capabilities Index | Not measured | 144.1 | Only Claude Opus 4.1 measured |
Agentic workDirectional onlyPublic boards, shown as published, kept out of the index3 boards measured
| Benchmark | UI-TARS 7B | Claude Opus 4.1 | Leader |
|---|---|---|---|
| METR time horizons | Not measured | 66.8% | Only Claude Opus 4.1 measured |
| GDPval | Not measured | 43.6% | Only Claude Opus 4.1 measured |
| DeepResearch Bench | Not measured | 48.3% | Only Claude Opus 4.1 measured |
| Neither model has a result on the other 39 boards here | |||
ReasoningDirectional onlyPublic boards, shown as published, kept out of the index11 boards measured
| Benchmark | UI-TARS 7B | Claude Opus 4.1 | Leader |
|---|---|---|---|
| GPQA Diamond | Not measured | 76.3% | Only Claude Opus 4.1 measured |
| MMMU Pro (Vals AI) | Not measured | 77.5% | Only Claude Opus 4.1 measured |
| GPQA Diamond (Epoch AI run) | Not measured | 77.3% | Only Claude Opus 4.1 measured |
| SimpleBench | Not measured | 60.0% | Only Claude Opus 4.1 measured |
| Mystery game puzzles | Not measured | 21.0% | Only Claude Opus 4.1 measured |
| EBR-bench | Not measured | 7.9% | Only Claude Opus 4.1 measured |
| VPCT | Not measured | 35.0% | Only Claude Opus 4.1 measured |
| LMArena Text: hard prompts | Not measured | 1442 | Only Claude Opus 4.1 measured |
| LMArena Text: expert | Not measured | 1438 | Only Claude Opus 4.1 measured |
| LMArena Text: hard prompts in English | Not measured | 1451 | Only Claude Opus 4.1 measured |
| Kagi LLM Benchmark | 20.7% | Not measured | Only UI-TARS 7B measured |
| Neither model has a result on the other 12 boards here | |||
MathDirectional onlyPublic boards, shown as published, kept out of the index8 boards measured
| Benchmark | UI-TARS 7B | Claude Opus 4.1 | Leader |
|---|---|---|---|
| AIME | Not measured | 78.2% | Only Claude Opus 4.1 measured |
| MATH 500 (Vals AI) | Not measured | 95.4% | Only Claude Opus 4.1 measured |
| MGSM (Vals AI) | Not measured | 94.4% | Only Claude Opus 4.1 measured |
| OTIS Mock AIME | Not measured | 68.9% | Only Claude Opus 4.1 measured |
| FrontierMath, tiers 1 to 3 | Not measured | 12.6% | Only Claude Opus 4.1 measured |
| FrontierMath, tier 4 | Not measured | 2.4% | Only Claude Opus 4.1 measured |
| LMArena Text: math | Not measured | 1431 | Only Claude Opus 4.1 measured |
| LMArena Text: mathematical work | Not measured | 1434 | Only Claude Opus 4.1 measured |
| Neither model has a result on the other 6 boards here | |||
KnowledgeDirectional onlyPublic boards, shown as published, kept out of the index13 boards measured
| Benchmark | UI-TARS 7B | Claude Opus 4.1 | Leader |
|---|---|---|---|
| MMLU Pro | Not measured | 87.9% | Only Claude Opus 4.1 measured |
| LegalBench (Vals AI) | Not measured | 83.5% | Only Claude Opus 4.1 measured |
| TaxEval (Vals AI) | Not measured | 73.7% | Only Claude Opus 4.1 measured |
| MortgageTax (Vals AI) | Not measured | 61.1% | Only Claude Opus 4.1 measured |
| MedQA (Vals AI) | Not measured | 93.6% | Only Claude Opus 4.1 measured |
| MedCode (Vals AI) | Not measured | 47.2% | Only Claude Opus 4.1 measured |
| LMArena Text: science | Not measured | 1422 | Only Claude Opus 4.1 measured |
| LMArena Text: medicine and healthcare | Not measured | 1416 | Only Claude Opus 4.1 measured |
| LMArena Text: legal and government | Not measured | 1421 | Only Claude Opus 4.1 measured |
| LMArena Text: business and finance | Not measured | 1406 | Only Claude Opus 4.1 measured |
| FACTS Benchmark Suite | Not measured | 40.4% | Only Claude Opus 4.1 measured |
| PRBench Finance | Not measured | 35.2% | Only Claude Opus 4.1 measured |
| PRBench Legal | Not measured | 34.0% | Only Claude Opus 4.1 measured |
| Neither model has a result on the other 7 boards here | |||
Web designDirectional onlyPublic boards, shown as published, kept out of the index8 boards measured
| Benchmark | UI-TARS 7B | Claude Opus 4.1 | Leader |
|---|---|---|---|
| Design Arena: websites | Not measured | 1176 | Only Claude Opus 4.1 measured |
| Design Arena: UI components | Not measured | 1162 | Only Claude Opus 4.1 measured |
| Design Arena: game dev | Not measured | 1171 | Only Claude Opus 4.1 measured |
| Design Arena: data viz | Not measured | 1166 | Only Claude Opus 4.1 measured |
| Design Arena: 3D | Not measured | 1152 | Only Claude Opus 4.1 measured |
| Design Arena: all code categories | Not measured | 1170 | Only Claude Opus 4.1 measured |
| Design Arena: SVG | Not measured | 1151 | Only Claude Opus 4.1 measured |
| Design Arena: ASCII art | Not measured | 1167 | Only Claude Opus 4.1 measured |
| Neither model has a result on the other 1 board here | |||
GeneralDirectional onlyPublic boards, shown as published, kept out of the index23 boards measured
| Benchmark | UI-TARS 7B | Claude Opus 4.1 | Leader |
|---|---|---|---|
| SAGE (Vals AI) | Not measured | 31.4% | Only Claude Opus 4.1 measured |
| MedScribe (Vals AI) | Not measured | 73.9% | Only Claude Opus 4.1 measured |
| DTBench | Not measured | 80.0% | Only Claude Opus 4.1 measured |
| LMCA | Not measured | 37.1% | Only Claude Opus 4.1 measured |
| Creative writing (Lech Mazur) | Not measured | 84.7% | Only Claude Opus 4.1 measured |
| LMArena Text | Not measured | 1419 | Only Claude Opus 4.1 measured |
| LMArena Text: instruction following | Not measured | 1435 | Only Claude Opus 4.1 measured |
| LMArena Text: creative writing | Not measured | 1412 | Only Claude Opus 4.1 measured |
| LMArena Text: English | Not measured | 1433 | Only Claude Opus 4.1 measured |
| LMArena Text: non-English | Not measured | 1405 | Only Claude Opus 4.1 measured |
| LMArena Text: Chinese | Not measured | 1427 | Only Claude Opus 4.1 measured |
| LMArena Text: German | Not measured | 1413 | Only Claude Opus 4.1 measured |
| LMArena Text: Japanese | Not measured | 1377 | Only Claude Opus 4.1 measured |
| LMArena Text: Korean | Not measured | 1381 | Only Claude Opus 4.1 measured |
| LMArena Text: French | Not measured | 1433 | Only Claude Opus 4.1 measured |
| LMArena Text: Spanish | Not measured | 1448 | Only Claude Opus 4.1 measured |
| LMArena Text: Russian | Not measured | 1422 | Only Claude Opus 4.1 measured |
| LMArena Text: Polish | Not measured | 1413 | Only Claude Opus 4.1 measured |
| LMArena Text: multi-turn | Not measured | 1444 | Only Claude Opus 4.1 measured |
| LMArena Text: longer queries | Not measured | 1455 | Only Claude Opus 4.1 measured |
| LMArena Text: writing and language | Not measured | 1418 | Only Claude Opus 4.1 measured |
| LMArena Text: entertainment and media | Not measured | 1398 | Only Claude Opus 4.1 measured |
| MultiChallenge | Not measured | 57.2% | Only Claude Opus 4.1 measured |
| Neither model has a result on the other 25 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 costs less
Repository review
50K input + 3K output tokens
UI-TARS 7B costs less
Agent loop
200K input + 10K output tokens
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 Opus 4.1 |
|---|---|---|
| Maker | ByteDance | Anthropic |
| Weights | Open weights | No public weights listed |
| Released | July 22, 2025 | August 5, 2025 |
| Context window | 128K | 200K |
| Input $/Mtok | $0.10 | $15.00 |
| Output $/Mtok | $0.20 | $75.00 |
| Coding index | Not benchmarked | Not benchmarked |
| Independent measurements behind it | 0 | 1 |
| Cost per solved task | Not computable | Not computable |
| Source | OpenRouter listing | OpenRouter listing |
Questions
Is UI-TARS 7B better than Claude Opus 4.1 for coding?
Neither model has a coding score yet, so the two cannot be ranked on coding. Pick UI-TARS 7B for output tokens 375.0x cheaper ($0.20 vs $75.00 per million).
Which is cheaper, UI-TARS 7B or Claude Opus 4.1?
UI-TARS 7B costs $0.10 per million input tokens and $0.20 per million output tokens; Claude Opus 4.1 costs $15.00 and $75.00. Cost per solved task is known for neither of them.
Which has the bigger context window, UI-TARS 7B or Claude Opus 4.1?
Claude Opus 4.1, with 200K tokens against 128K.
Has any benchmark tested both UI-TARS 7B and Claude Opus 4.1?
No benchmark has tested both yet. UI-TARS 7B has results on Kagi LLM Benchmark; Claude Opus 4.1 on LMArena WebDev, LiveCodeBench (Vals AI) and Terminal-Bench. Compare them on price and context instead. Data as of October 22, 2026.