UI-TARS 7B vs Claude Fable 5
UI-TARS 7B costs 250.0x less per output token than Claude Fable 5 ($0.20 vs $50.00 per million). Updated October 22, 2026; 1 benchmarks have a result for both.
vs boards won of 1 shared
UI-TARS 7B or Claude Fable 5: which to pick
Only Claude Fable 5 has a coding score, 79.3 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 250.0x cheaper ($0.20 vs $50.00 per million).
Model AUI-TARS 7BByteDanceNot benchmarkedNo coding benchmarkNo cost per solved taskModel BClaude Fable 5Anthropic79.3/1006 independent benchmarks$2.16 per solved task
- Shared results
- 1
- UI-TARS 7B only
- 0
- Claude Fable 5 only
- 134
- Like-for-like categories
- 1 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 Claude Fable 5 has a coding score, so no call is made.
One side not measured
Intelligence
AA Intelligence Index
No call
Only Claude Fable 5 has a AA Intelligence Index result, 49.6.
Directional only
Agentic work
AA Agentic Index
No call
Only Claude Fable 5 has a AA Agentic Index result, 50.7.
Directional only
Reasoning
GPQA Diamond
No call
Only Claude Fable 5 has a GPQA Diamond result, 93.2%.
Directional only
Knowledge
MMLU Pro
No call
Only Claude Fable 5 has a MMLU Pro result, 91.5%.
Directional only
Web design
Design Arena: websites
No call
Only Claude Fable 5 has a Design Arena: websites result, 1295.
Directional only
Long documents
Listed context window
Claude Fable 5
Claude Fable 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 Fable 5 $0.035 per turn.
List prices
Cost per solved task
One SWE-bench Verified task, solved
No call
Only Claude Fable 5 has the number, $2.16.
Directional only
At a glance
- Coding index
- UI-TARS 7B not benchmarked, Claude Fable 5 79.3 from 6 benchmarks.
- Widest category
- Reasoning, 70.7 points apart on average over 1 shared benchmark.
- Widest single gap
- Kagi LLM Benchmark: UI-TARS 7B 20.7% against Claude Fable 5 91.4%, 70.7 points apart on a 0 to 100 scale.
- Output price
- UI-TARS 7B $0.20 against Claude Fable 5 $50.00 per million output tokens, 250.0x apart.
- Context window
- UI-TARS 7B takes 128K tokens, Claude Fable 5 1M.
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 | UI-TARS 7B | Bars | Claude Fable 5 | Ties |
|---|---|---|---|---|
| Reasoning | 0 | 1 | ||
| All shared boards | 0 | 1 |
CodingDirectional onlySix of these boards feed the Vibe Coding Index26 boards, none shared
| Benchmark | UI-TARS 7B | Claude Fable 5 | Leader |
|---|---|---|---|
| LiveBench coding | Not measured | 98.3% | Only Claude Fable 5 measured |
| LMArena WebDev | Not measured | 1624 | Only Claude Fable 5 measured |
| AA Coding Index | Not measured | 76.5 | Only Claude Fable 5 measured |
| LiveCodeBench (Vals AI) | Not measured | 89.8% | Only Claude Fable 5 measured |
| SWE-bench Verified (Vals AI) | Not measured | 95.0% | Only Claude Fable 5 measured |
| Terminal-Bench 4 (Vals AI) | Not measured | 41.4% | Only Claude Fable 5 measured |
| Vibe Code Bench (Vals AI) | Not measured | 90.3% | Only Claude Fable 5 measured |
| Code migration (Vals AI) | Not measured | 55.1% | Only Claude Fable 5 measured |
| ProgramBench (Vals AI) | Not measured | 2.0% | Only Claude Fable 5 measured |
| Terminal-Bench 2.1 (Vals AI) | Not measured | 80.5% | Only Claude Fable 5 measured |
| SciCode | Not measured | 61.0% | Only Claude Fable 5 measured |
| Terminal-Bench Hard (Artificial Analysis run) | Not measured | 62.9% | Only Claude Fable 5 measured |
| FrontierCode | Not measured | 53.5% | Only Claude Fable 5 measured |
| FrontierSWE | Not measured | 47.0% | Only Claude Fable 5 measured |
| GSO-Bench | Not measured | 78.4% | Only Claude Fable 5 measured |
| WeirdML | Not measured | 91.9% | Only Claude Fable 5 measured |
| MirrorCode | Not measured | 63.9% | Only Claude Fable 5 measured |
| ALE-Bench | Not measured | 2041.3 | Only Claude Fable 5 measured |
| LMArena Text: coding | Not measured | 1518 | Only Claude Fable 5 measured |
| LMArena Text: software and IT | Not measured | 1513 | Only Claude Fable 5 measured |
| LMArena WebDev: React | Not measured | 1631 | Only Claude Fable 5 measured |
| LMArena WebDev: HTML | Not measured | 1651 | Only Claude Fable 5 measured |
| LMArena WebDev: from an image | Not measured | 1623 | Only Claude Fable 5 measured |
| SWE Atlas: Codebase QnA | Not measured | 39.0% | Only Claude Fable 5 measured |
| SWE Atlas: Refactoring | Not measured | 54.8% | Only Claude Fable 5 measured |
| SWE Atlas: Test Writing | Not measured | 55.6% | Only Claude Fable 5 measured |
| Neither model has a result on the other 17 boards here | |||
IntelligenceDirectional onlyPublic boards, shown as published, kept out of the index2 boards, none shared
| Benchmark | UI-TARS 7B | Claude Fable 5 | Leader |
|---|---|---|---|
| AA Intelligence Index | Not measured | 49.6 | Only Claude Fable 5 measured |
| Epoch Capabilities Index | Not measured | 163.4 | Only Claude Fable 5 measured |
Agentic workDirectional onlyPublic boards, shown as published, kept out of the index24 boards, none shared
| Benchmark | UI-TARS 7B | Claude Fable 5 | Leader |
|---|---|---|---|
| AA Agentic Index | Not measured | 50.7 | Only Claude Fable 5 measured |
| Design Arena: agents building web apps | Not measured | 1261 | Only Claude Fable 5 measured |
| Design Arena: agents building full-stack apps | Not measured | 1280 | Only Claude Fable 5 measured |
| Design Arena: agents building mobile apps | Not measured | 1263 | Only Claude Fable 5 measured |
| Design Arena: agents building Android apps | Not measured | 1283 | Only Claude Fable 5 measured |
| Design Arena: agents building Godot games | Not measured | 1292 | Only Claude Fable 5 measured |
| Design Arena: agents building games | Not measured | 1285 | Only Claude Fable 5 measured |
| Design Arena: HTML slides | Not measured | 1216 | Only Claude Fable 5 measured |
| Design Arena: PowerPoint slides | Not measured | 1269 | Only Claude Fable 5 measured |
| LiveBench research | Not measured | 86.9% | Only Claude Fable 5 measured |
| LiveBench browser | Not measured | 92.0% | Only Claude Fable 5 measured |
| LiveBench long agentic | Not measured | 53.2% | Only Claude Fable 5 measured |
| Finance Agent (Vals AI) | Not measured | 56.3% | Only Claude Fable 5 measured |
| Tax Agent Bench (Vals AI) | Not measured | 38.3% | Only Claude Fable 5 measured |
| Legal Research Bench (Vals AI) | Not measured | 49.5% | Only Claude Fable 5 measured |
| Excel Modeling Benchmark (Vals AI) | Not measured | 73.7% | Only Claude Fable 5 measured |
| Terminal-Bench-Science (Vals AI) | Not measured | 15.7% | Only Claude Fable 5 measured |
| τ²-Bench Telecom | Not measured | 98.5% | Only Claude Fable 5 measured |
| APEX-Agents | Not measured | 63.6% | Only Claude Fable 5 measured |
| Remote Labor Index | Not measured | 16.1% | Only Claude Fable 5 measured |
| PostTrainBench | Not measured | 41.8% | Only Claude Fable 5 measured |
| Vending-Bench 2 | Not measured | $5,680 | Only Claude Fable 5 measured |
| LMArena Search | Not measured | 1229 | Only Claude Fable 5 measured |
| MCP Atlas | Not measured | 83.3% | Only Claude Fable 5 measured |
| Neither model has a result on the other 18 boards here | |||
ReasoningLike-for-likePublic boards, shown as published, kept out of the index01
| Benchmark | UI-TARS 7B | Claude Fable 5 | Leader |
|---|---|---|---|
| GPQA Diamond | Not measured | 93.2% | Only Claude Fable 5 measured |
| LiveBench reasoning | Not measured | 89.7% | Only Claude Fable 5 measured |
| MMMU Pro (Vals AI) | Not measured | 89.3% | Only Claude Fable 5 measured |
| GPQA Diamond (Artificial Analysis run) | Not measured | 92.6% | Only Claude Fable 5 measured |
| Humanity's Last Exam (Artificial Analysis run) | Not measured | 55.5% | Only Claude Fable 5 measured |
| GPQA Diamond (Epoch AI run) | Not measured | 85.9% | Only Claude Fable 5 measured |
| ARC-AGI | Not measured | 98.5% | Only Claude Fable 5 measured |
| ARC-AGI-2 | Not measured | 89.2% | Only Claude Fable 5 measured |
| SimpleBench | Not measured | 81.9% | Only Claude Fable 5 measured |
| Chess puzzles | Not measured | 47.0% | Only Claude Fable 5 measured |
| Mystery game puzzles | Not measured | 52.0% | Only Claude Fable 5 measured |
| EBR-bench | Not measured | 39.5% | Only Claude Fable 5 measured |
| LMArena Text: hard prompts | Not measured | 1507 | Only Claude Fable 5 measured |
| LMArena Text: expert | Not measured | 1534 | Only Claude Fable 5 measured |
| LMArena Text: hard prompts in English | Not measured | 1509 | Only Claude Fable 5 measured |
| LMArena Vision: diagrams | Not measured | 1357 | Only Claude Fable 5 measured |
| LMArena Vision: homework | Not measured | 1361 | Only Claude Fable 5 measured |
| Kagi LLM Benchmark | 20.7% | 91.4% | Claude Fable 5 |
| Neither model has a result on the other 5 boards here | |||
MathDirectional onlyPublic boards, shown as published, kept out of the index7 boards, none shared
| Benchmark | UI-TARS 7B | Claude Fable 5 | Leader |
|---|---|---|---|
| OTIS Mock AIME | Not measured | 100.0% | Only Claude Fable 5 measured |
| FrontierMath, tiers 1 to 3 | Not measured | 87.0% | Only Claude Fable 5 measured |
| FrontierMath, tier 4 | Not measured | 90.2% | Only Claude Fable 5 measured |
| ProofBench | Not measured | 95.0% | Only Claude Fable 5 measured |
| LMArena Text: math | Not measured | 1516 | Only Claude Fable 5 measured |
| LMArena Text: mathematical work | Not measured | 1515 | Only Claude Fable 5 measured |
| MathArena | Not measured | 51.5% | Only Claude Fable 5 measured |
| Neither model has a result on the other 7 boards here | |||
KnowledgeDirectional onlyPublic boards, shown as published, kept out of the index13 boards, none shared
| Benchmark | UI-TARS 7B | Claude Fable 5 | Leader |
|---|---|---|---|
| MMLU Pro | Not measured | 91.5% | Only Claude Fable 5 measured |
| LegalBench (Vals AI) | Not measured | 88.6% | Only Claude Fable 5 measured |
| CorpFin (Vals AI) | Not measured | 71.8% | Only Claude Fable 5 measured |
| TaxEval (Vals AI) | Not measured | 76.9% | Only Claude Fable 5 measured |
| MortgageTax (Vals AI) | Not measured | 68.9% | Only Claude Fable 5 measured |
| MedCode (Vals AI) | Not measured | 56.1% | Only Claude Fable 5 measured |
| SimpleQA Verified | Not measured | 70.7% | Only Claude Fable 5 measured |
| LMArena Text: science | Not measured | 1506 | Only Claude Fable 5 measured |
| LMArena Text: medicine and healthcare | Not measured | 1480 | Only Claude Fable 5 measured |
| LMArena Text: legal and government | Not measured | 1498 | Only Claude Fable 5 measured |
| LMArena Text: business and finance | Not measured | 1480 | Only Claude Fable 5 measured |
| PRBench Finance | Not measured | 53.9% | Only Claude Fable 5 measured |
| PRBench Legal | Not measured | 52.6% | Only Claude Fable 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, none shared
| Benchmark | UI-TARS 7B | Claude Fable 5 | Leader |
|---|---|---|---|
| Design Arena: websites | Not measured | 1295 | Only Claude Fable 5 measured |
| Design Arena: UI components | Not measured | 1316 | Only Claude Fable 5 measured |
| Design Arena: game dev | Not measured | 1335 | Only Claude Fable 5 measured |
| Design Arena: data viz | Not measured | 1305 | Only Claude Fable 5 measured |
| Design Arena: 3D | Not measured | 1318 | Only Claude Fable 5 measured |
| Design Arena: all code categories | Not measured | 1310 | Only Claude Fable 5 measured |
| Design Arena: SVG | Not measured | 1301 | Only Claude Fable 5 measured |
| Design Arena: ASCII art | Not measured | 1329 | Only Claude Fable 5 measured |
| LiveBench design | Not measured | 84.7% | Only Claude Fable 5 measured |
GeneralDirectional onlyPublic boards, shown as published, kept out of the index36 boards, none shared
| Benchmark | UI-TARS 7B | Claude Fable 5 | Leader |
|---|---|---|---|
| LiveBench data analysis | Not measured | 80.5% | Only Claude Fable 5 measured |
| Public Benefits Bench (Vals AI) | Not measured | 70.4% | Only Claude Fable 5 measured |
| SAGE (Vals AI) | Not measured | 51.9% | Only Claude Fable 5 measured |
| MedScribe (Vals AI) | Not measured | 88.5% | Only Claude Fable 5 measured |
| Vals Multimodal Index | Not measured | 74.2% | Only Claude Fable 5 measured |
| IFBench | Not measured | 63.5% | Only Claude Fable 5 measured |
| Long context reasoning (AA-LCR) | Not measured | 82.3% | Only Claude Fable 5 measured |
| DTBench | Not measured | 98.4% | Only Claude Fable 5 measured |
| LMCA | Not measured | 61.2% | Only Claude Fable 5 measured |
| Surface Evolver Bench | Not measured | 95.0% | Only Claude Fable 5 measured |
| Furniture assembly | Not measured | 35.8% | Only Claude Fable 5 measured |
| LMArena Text | Not measured | 1491 | Only Claude Fable 5 measured |
| LMArena Text: instruction following | Not measured | 1502 | Only Claude Fable 5 measured |
| LMArena Text: creative writing | Not measured | 1494 | Only Claude Fable 5 measured |
| LMArena Vision | Not measured | 1325 | Only Claude Fable 5 measured |
| LMArena Text: English | Not measured | 1498 | Only Claude Fable 5 measured |
| LMArena Text: non-English | Not measured | 1482 | Only Claude Fable 5 measured |
| LMArena Text: Chinese | Not measured | 1543 | Only Claude Fable 5 measured |
| LMArena Text: German | Not measured | 1486 | Only Claude Fable 5 measured |
| LMArena Text: Japanese | Not measured | 1504 | Only Claude Fable 5 measured |
| LMArena Text: Korean | Not measured | 1487 | Only Claude Fable 5 measured |
| LMArena Text: French | Not measured | 1505 | Only Claude Fable 5 measured |
| LMArena Text: Spanish | Not measured | 1498 | Only Claude Fable 5 measured |
| LMArena Text: Russian | Not measured | 1503 | Only Claude Fable 5 measured |
| LMArena Text: Polish | Not measured | 1494 | Only Claude Fable 5 measured |
| LMArena Text: multi-turn | Not measured | 1504 | Only Claude Fable 5 measured |
| LMArena Text: longer queries | Not measured | 1510 | Only Claude Fable 5 measured |
| LMArena Text: writing and language | Not measured | 1501 | Only Claude Fable 5 measured |
| LMArena Text: entertainment and media | Not measured | 1482 | Only Claude Fable 5 measured |
| LMArena Vision: English | Not measured | 1323 | Only Claude Fable 5 measured |
| LMArena Vision: Chinese | Not measured | 1373 | Only Claude Fable 5 measured |
| LMArena Vision: reading text | Not measured | 1337 | Only Claude Fable 5 measured |
| LMArena Vision: humor | Not measured | 1341 | Only Claude Fable 5 measured |
| LMArena Vision: creative writing | Not measured | 1338 | Only Claude Fable 5 measured |
| LMArena Document | Not measured | 1496 | Only Claude Fable 5 measured |
| EQ-Bench 4 | Not measured | 1340 | Only Claude Fable 5 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 | UI-TARS 7B | Claude Fable 5 | 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
UI-TARS 7B$0.0002Fits in one request
Claude Fable 5$0.035Fits in one request
UI-TARS 7B costs less
Repository review
50K input + 3K output tokens
UI-TARS 7B$0.0056Fits in one request
Claude Fable 5$0.650Fits in one request
UI-TARS 7B costs less
Agent loop
200K input + 10K output tokens
UI-TARS 7B$0.022Exceeds the context window
Claude Fable 5$2.50Fits 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 Fable 5 |
|---|---|---|
| Maker | ByteDance | Anthropic |
| Weights | Open weights | No public weights listed |
| Released | July 22, 2025 | June 9, 2026 |
| Context window | 128K | 1M |
| Input $/Mtok | $0.10 | $10.00 |
| Output $/Mtok | $0.20 | $50.00 |
| Coding index | Not benchmarked | 79.3 |
| Independent measurements behind it | 0 | 6 |
| Cost per solved task | Not computable | $2.16 |
| Source | OpenRouter listing | OpenRouter listing |
Questions
Is UI-TARS 7B better than Claude Fable 5 for coding?
Only Claude Fable 5 has a coding score, 79.3 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 250.0x cheaper ($0.20 vs $50.00 per million).
Which is cheaper, UI-TARS 7B or Claude Fable 5?
UI-TARS 7B costs $0.10 per million input tokens and $0.20 per million output tokens; Claude Fable 5 costs $10.00 and $50.00. Cost per solved task is known for Claude Fable 5 only.
Which has the bigger context window, UI-TARS 7B or Claude Fable 5?
Claude Fable 5, with 1M tokens against 128K.
How many benchmarks is the UI-TARS 7B vs Claude Fable 5 comparison based on?
1 benchmarks have a result for both models as of October 22, 2026. UI-TARS 7B's coding index rests on 0 independent benchmarks and Claude Fable 5's on 6. Where only one model has a result, the page says so instead of naming a winner.