Compare · ModelsLive · 3 picked · head to head
Claude Fable 5 vs Gemini 4 Carbon vs GPT-6 Astra
Side by side · benchmarks, pricing, and signals you can act on.
Winner summary
GPT-6 Astra wins on 14/26 benchmarks
GPT-6 Astra wins 14 of 26 shared benchmarks. Leads in general · math · coding.
Category leads
agentic·Gemini 4 Carbonreasoning·Claude Fable 5knowledge·Gemini 4 Carbongeneral·GPT-6 Astramath·GPT-6 Astracoding·GPT-6 Astraspeed·Claude Fable 5
Hype vs Reality
Attention vs performance
Claude Fable 5
#30 by perf·#5 by attention
Gemini 4 Carbon
#1 by perf·#8 by attention
GPT-6 Astra
#9 by perf·#2 by attention
Best value
Gemini 4 Carbon
5.7x better value than GPT-6 Astra
Claude Fable 5
2.3 pts/$
$30.00/M
Gemini 4 Carbon
14.3 pts/$
$6.00/M
GPT-6 Astra
2.5 pts/$
$30.00/M
Vendor risk
Who is behind the model
Anthropic
$965.0B·Tier 1
Google DeepMind
$3.61T·Tier 1
OpenAI
$840.0B·Tier 1
Head to head
26 benchmarks · 3 models
Claude Fable 5Gemini 4 CarbonGPT-6 Astra
APEX-Agents
Gemini 4 Carbon leads by +19.5
APEX-Agents · evaluates AI agents on complex, multi-step tasks requiring planning, tool use, and autonomous decision-making in realistic environments.
Claude Fable 5
63.6
Gemini 4 Carbon
84.2
GPT-6 Astra
64.7
ARC-AGI
ARC-AGI · the original Abstraction and Reasoning Corpus, testing whether AI can solve novel visual pattern recognition tasks without memorization.
Claude Fable 5
98.5
Gemini 4 Carbon
98.5
GPT-6 Astra
98.5
ARC-AGI-2
GPT-6 Astra leads by +0.8
ARC-AGI-2 · the second iteration of the Abstraction and Reasoning Corpus, testing novel pattern recognition and abstract reasoning without prior training data.
Claude Fable 5
89.2
Gemini 4 Carbon
94.2
GPT-6 Astra
95.0
Chess Puzzles
Gemini 4 Carbon leads by +2.1
Chess Puzzles · tests strategic and tactical reasoning by having models solve chess puzzle positions, evaluating lookahead and pattern recognition abilities.
Claude Fable 5
44.2
Gemini 4 Carbon
76.8
GPT-6 Astra
74.8
Ebr Bench
Gemini 4 Carbon leads by +11.4
Claude Fable 5
39.5
Gemini 4 Carbon
87.6
GPT-6 Astra
76.2
FrontierMath-Tier-4-v2-Private
GPT-6 Astra leads by +2.5
Claude Fable 5
90.2
Gemini 4 Carbon
95.1
GPT-6 Astra
97.6
FrontierMath-Tiers-1-3-v2-Private
GPT-6 Astra leads by +3.3
Claude Fable 5
91.6
Gemini 4 Carbon
95.3
GPT-6 Astra
98.6
Frontierswe
GPT-6 Astra leads by +4.3
Claude Fable 5
47.0
Gemini 4 Carbon
61.2
GPT-6 Astra
65.5
Furniture Assembly
GPT-6 Astra leads by +3.6
Claude Fable 5
8.3
Gemini 4 Carbon
67.9
GPT-6 Astra
71.4
GPQA diamond
Gemini 4 Carbon leads by +1.0
Graduate-Level Google-Proof QA (Diamond set) · expert-crafted questions in physics, biology, and chemistry that are difficult even for domain PhDs.
Claude Fable 5
84.5
Gemini 4 Carbon
99.3
GPT-6 Astra
98.3
Mystery Game Puzzles
GPT-6 Astra leads by +12.1
Claude Fable 5
47.1
Gemini 4 Carbon
70.3
GPT-6 Astra
82.4
OTIS Mock AIME 2024-2025
OTIS Mock AIME 2024-2025 · simulated American Invitational Mathematics Examination problems testing advanced problem-solving skills.
Claude Fable 5
100.0
Gemini 4 Carbon
100.0
GPT-6 Astra
100.0
SimpleBench
Gemini 4 Carbon leads by +7.3
SimpleBench · tests fundamental reasoning capabilities with straightforward problems designed to expose gaps in basic logical and spatial thinking.
Claude Fable 5
78.3
Gemini 4 Carbon
91.1
GPT-6 Astra
83.8
SimpleQA Verified
Gemini 4 Carbon leads by +2.1
SimpleQA Verified · short factual questions with verified answers, measuring factual accuracy and the tendency to hallucinate or provide incorrect information.
Claude Fable 5
70.7
Gemini 4 Carbon
77.7
GPT-6 Astra
75.6
Artificial Analysis · Quality Index
Claude Fable 5 leads by +7.2
Claude Fable 5
59.9
GPT-6 Astra
52.7
Deepswe
GPT-6 Astra leads by +4.2
Claude Fable 5
69.9
GPT-6 Astra
74.1
Dtbench
Claude Fable 5 leads by +1.8
Claude Fable 5
97.3
GPT-6 Astra
95.5
Frontiercode
Claude Fable 5 leads by +0.2
Claude Fable 5
53.5
GPT-6 Astra
53.3
GSO-Bench
GPT-6 Astra leads by +1.0
GSO-Bench · evaluates AI models on real-world open-source software engineering tasks, testing the ability to understand and resolve actual GitHub issues.
Claude Fable 5
78.4
GPT-6 Astra
79.4
Lmca
GPT-6 Astra leads by +3.8
Claude Fable 5
71.9
GPT-6 Astra
75.7
Mirrorcode
Claude Fable 5 leads by +10.0
Claude Fable 5
70.0
GPT-6 Astra
60.0
PostTrainBench
GPT-6 Astra leads by +4.6
Claude Fable 5
37.3
GPT-6 Astra
41.9
Proofbench
GPT-6 Astra leads by +4.0
Claude Fable 5
95.0
GPT-6 Astra
99.0
Remote Labor Index
GPT-6 Astra leads by +4.7
Claude Fable 5
16.1
GPT-6 Astra
20.8
WeirdML
GPT-6 Astra leads by +1.6
WeirdML · tests models on unusual and adversarial machine learning tasks that require creative problem-solving beyond standard patterns.
Claude Fable 5
91.9
GPT-6 Astra
93.6
WeirdML v3
GPT-6 Astra leads by +29.1
Claude Fable 5
31.0
GPT-6 Astra
60.0
Full benchmark table
| Benchmark | Claude Fable 5 | Gemini 4 Carbon | GPT-6 Astra |
|---|---|---|---|
APEX-Agents APEX-Agents · evaluates AI agents on complex, multi-step tasks requiring planning, tool use, and autonomous decision-making in realistic environments. | 63.6 | 84.2 | 64.7 |
ARC-AGI ARC-AGI · the original Abstraction and Reasoning Corpus, testing whether AI can solve novel visual pattern recognition tasks without memorization. | 98.5 | 98.5 | 98.5 |
ARC-AGI-2 ARC-AGI-2 · the second iteration of the Abstraction and Reasoning Corpus, testing novel pattern recognition and abstract reasoning without prior training data. | 89.2 | 94.2 | 95.0 |
Chess Puzzles Chess Puzzles · tests strategic and tactical reasoning by having models solve chess puzzle positions, evaluating lookahead and pattern recognition abilities. | 44.2 | 76.8 | 74.8 |
Ebr Bench | 39.5 | 87.6 | 76.2 |
FrontierMath-Tier-4-v2-Private | 90.2 | 95.1 | 97.6 |
FrontierMath-Tiers-1-3-v2-Private | 91.6 | 95.3 | 98.6 |
Frontierswe | 47.0 | 61.2 | 65.5 |
Furniture Assembly | 8.3 | 67.9 | 71.4 |
GPQA diamond Graduate-Level Google-Proof QA (Diamond set) · expert-crafted questions in physics, biology, and chemistry that are difficult even for domain PhDs. | 84.5 | 99.3 | 98.3 |
Mystery Game Puzzles | 47.1 | 70.3 | 82.4 |
OTIS Mock AIME 2024-2025 OTIS Mock AIME 2024-2025 · simulated American Invitational Mathematics Examination problems testing advanced problem-solving skills. | 100.0 | 100.0 | 100.0 |
SimpleBench SimpleBench · tests fundamental reasoning capabilities with straightforward problems designed to expose gaps in basic logical and spatial thinking. | 78.3 | 91.1 | 83.8 |
SimpleQA Verified SimpleQA Verified · short factual questions with verified answers, measuring factual accuracy and the tendency to hallucinate or provide incorrect information. | 70.7 | 77.7 | 75.6 |
Artificial Analysis · Quality Index | 59.9 | n/a | 52.7 |
Deepswe | 69.9 | n/a | 74.1 |
Dtbench | 97.3 | n/a | 95.5 |
Frontiercode | 53.5 | n/a | 53.3 |
GSO-Bench GSO-Bench · evaluates AI models on real-world open-source software engineering tasks, testing the ability to understand and resolve actual GitHub issues. | 78.4 | n/a | 79.4 |
Lmca | 71.9 | n/a | 75.7 |
Mirrorcode | 70.0 | n/a | 60.0 |
PostTrainBench | 37.3 | n/a | 41.9 |
Proofbench | 95.0 | n/a | 99.0 |
Remote Labor Index | 16.1 | n/a | 20.8 |
WeirdML WeirdML · tests models on unusual and adversarial machine learning tasks that require creative problem-solving beyond standard patterns. | 91.9 | n/a | 93.6 |
WeirdML v3 | 31.0 | n/a | 60.0 |
Pricing · per 1M tokens · projected $/mo at 10M tokens
| Model | Input | Output | Context | Projected $/mo |
|---|---|---|---|---|
| $10.00 | $50.00 | 1.0M tokens (~500 books) | $200.00 | |
| $2.00 | $10.00 | 1.0M tokens (~524 books) | $40.00 | |
| $10.00 | $50.00 | 1.1M tokens (~525 books) | $200.00 |
People also compared