Fable-5 vs Kolibri-1
New ComparisonScores are fetched live from the current database on every visit — nothing is cached.
Winner
Fable-5
Fable-5 wins this comparison with an overall score of 89 vs 37 for Kolibri-1, leading 3 of 3 shared benchmarks.
Fable-5
89.2
Kolibri-1
37.2
Based on 3 of 85 benchmarks scored by all 2 selected models.
Benchmark Scores Comparison
Fable-5
Kolibri-1
Fable-5
Score: 15.0
Kolibri-1
78.1B params
Coverage matrix
2 models × 85 benchmarks — hover a benchmark name for its category
Shading = score relative to the best score among the selected models:
Low
Mid
High
Missing (—)
The totals row shows how many of the 85 benchmarks each model has scores for.
| Benchmark | Kolibri-1 Aleph Alpha | Fable-5 Anthropic |
|---|---|---|
| OmniScience Accuracy | 14.8 | — |
| GPQA Diamond | 84.3 | 92.6 |
| IFBench (prompt loose) | 78.1 | — |
| SealQA (12 distractors, <24k) | 61 | — |
| LiveCodeBench v6 | 85.9 | — |
| Automation-Bench | — | 29.1 |
| SWE-bench Pro | — | 80 |
| BFCL-V4 | 61.4 | — |
| PRBench-Finance | — | 55.8 |
| AA-LCR | 68.3 | — |
| AIME 26 | 96 | — |
| Agents' Last Exam | — | 23.8 |
| Toolathlon Verified | — | 77.9 |
| PLawBench | — | 70.2 |
| FRAMES (<24k) | 71.2 | — |
| German Public Sector | 75 | — |
| OmniScience Non-Hallucination | 44 | — |
| Aerospace | 58.9 | — |
| Agentic Wiki QA (DE) | 69.4 | — |
| QwenReactBench | — | 1770 |
| AIME 2025 (DE) | 87.5 | — |
| Overall (EN) | 75.5 | — |
| QwenSVGBench | — | 1690 |
| MuSiQue (EN) | 77.3 | — |
| SealQA (no distractors, >24k) | 100 | — |
| PRBench-Legal | — | 57.6 |
| PostTrainBench | — | 41.8 |
| DSBench-FullStack | — | 77.2 |
| Terminal-Bench 3.0 | — | 33.7 |
| HumanEval+ | 92.7 | — |
| MMLU-ProX | 75.5 | — |
| WorkSpaceBench | — | 68.7 |
| ExploitGym (6h) | — | 247 |
| HLE (with tools) | — | 64.5 |
| FrontierSWE | — | 88.2 |
| BrowseComp | 29.4 | — |
| SkillsBench | — | 70.9 |
| BFCL v4 (web search) | 62.5 | — |
| RGB Closed-Book | 51 | — |
| CoWorkBench | — | 75.9 |
| Industrial Drive Technology | 60 | — |
| AA-Omniscience Index (public set) | -32.8 | — |
| PaperBench | — | 88.8 |
| Humanity's Last Exam (DE) | 15.9 | — |
| IFBench | — | 63.5 |
| SWE-bench Verified | 66.4 | — |
| Humanity's Last Exam | 21.5 | 53.3 |
| BFCL v4 (non-live AST) | 79.1 | — |
| BFCL v4 (live) | 78.9 | — |
| DSBench-Hard | — | 68.3 |
| BFCL v4 (memory) | 62.8 | — |
| AIME 2026 (DE) | 90 | — |
| QwenSWEBench | — | 86.3 |
| ExploitGym (2h) | — | 181 |
| Automotive Supplier | 99 | — |
| DeepSWE | — | 70 |
| GPQA Diamond (DE) | 81.3 | — |
| ExploitBench | — | 78 |
| SQuAD (Utility Accuracy) | 83.4 | — |
| Terminal Bench 2.1 | 27.7 | 84.6 |
| AIME 2025 | 96.9 | — |
| BFCL v4 (multi-turn) | 47.5 | — |
| RGB Fact-Check (Error Correction) | 34 | — |
| JobBench | — | 57.4 |
| $OneMillion-Bench | — | 55.9 |
| SQuAD (M/A Grounding Score) | 23.4 | — |
| LongBenchPro | 64.5 | — |
| QwenQoderBench | — | 63.1 |
| FRAMES (>24k) | 73 | — |
| Semiconductors | 80.4 | — |
| Honeypot | 80.8 | — |
| WideSearch | — | 81.2 |
| TauBench V3 Banking | 38.1 | — |
| ProgramBench | — | 33 |
| AndroidBench | — | 84.5 |
| SWE-Marathon | — | 33.1 |
| Cybergym | — | 83.1 |
| DeepSWE 1.1 | — | 70 |
| RGB Negative (Abstention) | 85.6 | — |
| MMLU Pro COT | 80 | — |
| Overall (DE) | 70.8 | — |
| MLS-Bench-Lite | — | 49.9 |
| SealQA (12 distractors, >24k) | 65.1 | — |
| SealQA (no distractors, <24k) | 80.7 | — |
| GDPVal-AA v2 | — | 1743 |
| Covered | 49/85 | 39/85 |
Scorecard
Overall 0–100 score — category strengths (50%), shared-benchmark wins (30%), coverage (20%)
Category strengths
Where each model stands, per benchmark category — score relative to the best model, adjusted for benchmark coverage. This also ranks models when they share few benchmarks.
| Category | Fable-5 | Kolibri-1 |
|---|---|---|
| Agentic Coding 19 benchmarks | 100 | 32.4 |
| STEM & Reasoning 12 benchmarks | 100 | 60.8 |
Fully comparable benchmarks
Scored by all 2 selected models (3)
| Benchmark | Category | Fable-5 | Kolibri-1 |
|---|---|---|---|
| GPQA Diamond | stem_reasoning | 92.6 | 84.3 |
| Humanity's Last Exam | stem_reasoning | 53.3 | 21.5 |
| Terminal Bench 2.1 | coding_agent | 84.6 | 27.7 |
Partial coverage benchmarks
Scored by at least one but not all selected models (82) — missing scores show as —
| Benchmark | Category | Fable-5 | Kolibri-1 |
|---|---|---|---|
| OmniScience Accuracy 1/2 models | knowledge | — | 14.8 |
| IFBench (prompt loose) 1/2 models | instruction_following | — | 78.1 |
| SealQA (12 distractors, <24k) 1/2 models | agentic | — | 61 |
| LiveCodeBench v6 1/2 models | stem_reasoning | — | 85.9 |
| Automation-Bench 1/2 models | general_agent | 29.1 | — |
| SWE-bench Pro 1/2 models | coding_agent | 80 | — |
| BFCL-V4 1/2 models | general_agent | — | 61.4 |
| PRBench-Finance 1/2 models | general_capabilities | 55.8 | — |
| AA-LCR 1/2 models | long_context | — | 68.3 |
| AIME 26 1/2 models | stem_reasoning | — | 96 |
| Agents' Last Exam 1/2 models | general_agent | 23.8 | — |
| Toolathlon Verified 1/2 models | general_agent | 77.9 | — |
| PLawBench 1/2 models | general_capabilities | 70.2 | — |
| FRAMES (<24k) 1/2 models | agentic | — | 71.2 |
| German Public Sector 1/2 models | agentic | — | 75 |
| OmniScience Non-Hallucination 1/2 models | knowledge | — | 44 |
| Aerospace 1/2 models | agentic | — | 58.9 |
| Agentic Wiki QA (DE) 1/2 models | agentic | — | 69.4 |
| QwenReactBench 1/2 models | coding_agent | 1770 | — |
| AIME 2025 (DE) 1/2 models | stem_reasoning | — | 87.5 |
| Overall (EN) 1/2 models | stem_reasoning | — | 75.5 |
| QwenSVGBench 1/2 models | coding_agent | 1690 | — |
| MuSiQue (EN) 1/2 models | agentic | — | 77.3 |
| SealQA (no distractors, >24k) 1/2 models | agentic | — | 100 |
| PRBench-Legal 1/2 models | general_capabilities | 57.6 | — |
| PostTrainBench 1/2 models | coding_agent | 41.8 | — |
| DSBench-FullStack 1/2 models | coding_agent | 77.2 | — |
| Terminal-Bench 3.0 1/2 models | coding_agent | 33.7 | — |
| HumanEval+ 1/2 models | code | — | 92.7 |
| MMLU-ProX 1/2 models | multilingual | — | 75.5 |
| WorkSpaceBench 1/2 models | general_agent | 68.7 | — |
| ExploitGym (6h) 1/2 models | cybersecurity | 247 | — |
| HLE (with tools) 1/2 models | stem_reasoning | 64.5 | — |
| FrontierSWE 1/2 models | coding_agent | 88.2 | — |
| BrowseComp 1/2 models | general_agent | — | 29.4 |
| SkillsBench 1/2 models | general_agent | 70.9 | — |
| BFCL v4 (web search) 1/2 models | agentic | — | 62.5 |
| RGB Closed-Book 1/2 models | grounding_hallucination | — | 51 |
| CoWorkBench 1/2 models | general_agent | 75.9 | — |
| Industrial Drive Technology 1/2 models | agentic | — | 60 |
| AA-Omniscience Index (public set) 1/2 models | knowledge | — | -32.8 |
| PaperBench 1/2 models | coding_agent | 88.8 | — |
| Humanity's Last Exam (DE) 1/2 models | stem_reasoning | — | 15.9 |
| IFBench 1/2 models | instruction_following | 63.5 | — |
| SWE-bench Verified 1/2 models | coding_agent | — | 66.4 |
| BFCL v4 (non-live AST) 1/2 models | agentic | — | 79.1 |
| BFCL v4 (live) 1/2 models | agentic | — | 78.9 |
| DSBench-Hard 1/2 models | coding_agent | 68.3 | — |
| BFCL v4 (memory) 1/2 models | agentic | — | 62.8 |
| AIME 2026 (DE) 1/2 models | stem_reasoning | — | 90 |
| QwenSWEBench 1/2 models | coding_agent | 86.3 | — |
| ExploitGym (2h) 1/2 models | cybersecurity | 181 | — |
| Automotive Supplier 1/2 models | agentic | — | 99 |
| DeepSWE 1/2 models | coding_agent | 70 | — |
| GPQA Diamond (DE) 1/2 models | stem_reasoning | — | 81.3 |
| ExploitBench 1/2 models | cybersecurity | 78 | — |
| SQuAD (Utility Accuracy) 1/2 models | grounding_hallucination | — | 83.4 |
| AIME 2025 1/2 models | stem_reasoning | — | 96.9 |
| BFCL v4 (multi-turn) 1/2 models | agentic | — | 47.5 |
| RGB Fact-Check (Error Correction) 1/2 models | grounding_hallucination | — | 34 |
| JobBench 1/2 models | general_agent | 57.4 | — |
| $OneMillion-Bench 1/2 models | general_capabilities | 55.9 | — |
| SQuAD (M/A Grounding Score) 1/2 models | grounding_hallucination | — | 23.4 |
| LongBenchPro 1/2 models | long_context | — | 64.5 |
| QwenQoderBench 1/2 models | coding_agent | 63.1 | — |
| FRAMES (>24k) 1/2 models | agentic | — | 73 |
| Semiconductors 1/2 models | agentic | — | 80.4 |
| Honeypot 1/2 models | agentic | — | 80.8 |
| WideSearch 1/2 models | general_agent | 81.2 | — |
| TauBench V3 Banking 1/2 models | general_agent | — | 38.1 |
| ProgramBench 1/2 models | coding_agent | 33 | — |
| AndroidBench 1/2 models | coding_agent | 84.5 | — |
| SWE-Marathon 1/2 models | coding_agent | 33.1 | — |
| Cybergym 1/2 models | general_agent | 83.1 | — |
| DeepSWE 1.1 1/2 models | coding_agent | 70 | — |
| RGB Negative (Abstention) 1/2 models | grounding_hallucination | — | 85.6 |
| MMLU Pro COT 1/2 models | knowledge | — | 80 |
| Overall (DE) 1/2 models | stem_reasoning | — | 70.8 |
| MLS-Bench-Lite 1/2 models | coding_agent | 49.9 | — |
| SealQA (12 distractors, >24k) 1/2 models | agentic | — | 65.1 |
| SealQA (no distractors, <24k) 1/2 models | agentic | — | 80.7 |
| GDPVal-AA v2 1/2 models | general_agent | 1743 | — |
Comparison FAQ
Which model is better, Fable-5 or Kolibri-1?
Fable-5 wins this comparison with an overall score of 89 vs 37 for Kolibri-1, leading 3 of 3 shared benchmarks.
Which model is better at STEM & Reasoning, Fable-5 or Kolibri-1?
Fable-5 leads the STEM & Reasoning category (strength score 100/100 across 2 comparable benchmarks).
How reliable is this comparison verdict?
Based on 3 of 85 benchmarks scored by all 2 selected models.