Updated October 17, 2026. Rank says GPT-6 Astra is ahead. Price, access, and your workload can each overturn that. Public scores include evidence status and uncertainty.
GPT-6 Astra has the higher public score, 83.88 versus 55.89, and the 90% score intervals do not overlap. 8 results are shared. Category rows resting on Estimated evidence or different benchmark sets are marked directional and do not name a winner.
Recommendations appear only when a shared evidence basis or an explicit operating constraint supports the call. Secondary and unsupported use cases stay disclosed below the initial list.
Code generation, repair, and software-engineering tasks
GPT-6 Astra
GPT-6 Astra has the higher public coding point estimate, 75.5 to 44.7, with Supported evidence for both models. The conditional ranges do not establish rank confidence.
Tool use, computer use, and multi-step task completion
GPT-6 Astra
GPT-6 Astra has the higher public agentic point estimate, 74.8 to 46.3, with Supported evidence for both models. The conditional ranges do not establish rank confidence.
Prompts that approach the documented context limit
GPT-6 Astra
GPT-6 Astra has the larger documented context window.
1K fresh input + 500 output tokens
Claude Sonnet 4.6
Claude Sonnet 4.6 has the lower estimated token cost for this stated workload. Costs use the listed standard API rates.
50K fresh input + 3K output tokens
Claude Sonnet 4.6
Claude Sonnet 4.6 has the lower estimated token cost for this stated workload. Costs use the listed standard API rates.
200K cached + 20K fresh input + 10K output tokens
Not enough matched evidence
The page does not recommend a cost winner because at least one model cannot fit the stated workload in one request. Claude Sonnet 4.6 does not fit this workload in one request. Claude Sonnet 4.6 has no published cached-input rate, so cached tokens use its listed input rate.
Choose a job from the LLM Selector's task catalog to see the category row it rests on, under the same basis rules as the table below. A directional row stays directional; choosing a job never creates a winner.
The same task catalog as the LLM Selector. Each job names the evidence surface it rests on; nothing here adds a new score.
Like-for-like · BenchAlign v5.8
GPT-6 Astra has the higher coding point estimate. Conditional score ranges do not establish rank confidence.
Coding scores combine specific tasks and setups. Match the editor, harness, and effort to your workflow.
Shared results can support a head-to-head reading. Results present for only one model describe coverage, not superiority.
Only shared public evidence is shown. Sparse evidence stays a ruled list rather than being closed into a radar shape.
Too few matched category axes support a radar. The ruled list below shows only shared benchmark results; positions use each benchmark’s normalized display scale when available.
OSWorld 2.0Agentic
Normalized gap 64.3ARC-AGI-2Reasoning
Normalized gap 36.7GPQAKnowledge
Normalized gap 6.1Each row shows the public-lane category score for both models: the BenchAlign v5.8 lane for agentic, coding, and knowledge, and the weighted public rows elsewhere, with the same rank each model page reports. A row is like-for-like only when both scores rest on Supported evidence or the same weighted set. Directional and not-comparable rows remain visible, but they never receive a winner.
| Category | Claude Sonnet 4.6 | GPT-6 Astra | Basis | Reading |
|---|---|---|---|---|
| Agentic | 46.3Supported · #55/128 | 74.8Supported · #7/128 | Like-for-likeBenchAlign v5.8 lane · 10 vs 14 public rows | GPT-6 Astra leads |
| Coding | 44.7Supported · #54/150 | 75.5Supported · #4/150 | Like-for-likeBenchAlign v5.8 lane · 9 vs 6 public rows | GPT-6 Astra leads |
| Knowledge | 55.4Supported · #58/184 | 86.2Supported · #3/184 | Like-for-likeBenchAlign v5.8 lane · 6 vs 7 public rows | GPT-6 Astra leads |
| Reasoning | 62.8Unranked · 4 rankable rows | 93.3#1/28 | Not comparableProvisional lane · 1 vs 2 weighted rows | Not comparable |
| Multimodal | 58.7#45/54 | 87.3Unranked · 3 rankable rows | Not comparableProvisional lane · 1 vs 0 weighted rows | Not comparable |
| Multilingual | Not ranked | Not ranked | Not comparableProvisional lane · 0 vs 0 weighted rows | Not comparable |
| Instruction following | 48.0#99/148 | Not ranked | Not comparableProvisional lane · 0 vs 0 weighted rows | Not comparable |
| Math | Not ranked | 97.3Unranked · 2 rankable rows | Not comparableProvisional lane · 0 vs 2 weighted rows | Not comparable |
Ranks count the models scored in each category’s lane, so the agentic, coding, and knowledge denominators (BenchAlign v5.8) differ from the provisional-lane categories. Unranked scores sit on the provisional lane’s scale but never name a winner. Like-for-like rows within 0.5 points read as a practical tie.
Three fixed token mixes turn per-token rates into comparable decisions. Each scenario states context fit and whether cached input had to fall back to the published list-input rate.
1K fresh input + 500 output tokens
Claude Sonnet 4.6 has the lower modeled cost
Costs use the listed standard API rates.
50K fresh input + 3K output tokens
Claude Sonnet 4.6 has the lower modeled cost
Costs use the listed standard API rates.
200K cached + 20K fresh input + 10K output tokens
Claude Sonnet 4.6 does not fit this workload in one request. Claude Sonnet 4.6 has no published cached-input rate, so cached tokens use its listed input rate.
Sourced differences are shown directly. Missing facts stay explicit instead of being inferred from a model name or family.
Maximum documented context; output-token limits may be lower.
Claude Sonnet 4.6
200K
GPT-6 Astra
Claude Sonnet 4.6
Not sourced
GPT-6 Astra
gpt-6-astra
OpenAI GPT-6 Astra model documentationA missing cached-input rate falls back to the listed input rate only in the stated workload estimate.
Claude Sonnet 4.6
Not published
GPT-6 Astra
$1 per 1M cached input tokens
OpenAI GPT-6 Astra model documentationClaude Sonnet 4.6
Not sourced
GPT-6 Astra
text, image
OpenAI GPT-6 Astra model documentationClaude Sonnet 4.6
Not sourced
GPT-6 Astra
Claude Sonnet 4.6
Not sourced
GPT-6 Astra
Limited Availability · OpenAI Responses API, OpenAI Chat Completions API, ChatGPT Plus, Pro, Business, and Enterprise, Microsoft Azure, AWS Bedrock
OpenAI GPT-6 Astra model documentationClaude Sonnet 4.6
Non-Reasoning
GPT-6 Astra
Reasoning
Claude Sonnet 4.6
Proprietary
GPT-6 Astra
Proprietary
Claude Sonnet 4.6
Proprietary
GPT-6 Astra
Proprietary
Claude Sonnet 4.6
2026-02-01
GPT-6 Astra
2026-09-03
GPT-6 Astra has the higher public score, 83.88 versus 55.89, and the 90% score intervals do not overlap. That is the clearest overall quality signal in the current public data.
GPT-6 Astra has the higher public coding point estimate, 75.5 to 44.7, with Supported evidence for both models. The conditional ranges do not establish rank confidence.
GPT-6 Astra has the higher public agentic tasks point estimate, 74.8 to 46.3, with Supported evidence for both models. The conditional ranges do not establish rank confidence.
For the stated presets, chat costs $0.0105 on Claude Sonnet 4.6 and $0.035 on GPT-6 Astra; repository review costs $0.195 and $0.65; the cache-heavy agent loop costs $0.81 and $0.9. Claude Sonnet 4.6 does not fit this workload in one request. Claude Sonnet 4.6 has no published cached-input rate, so cached tokens use its listed input rate.
GPT-6 Astra has the larger documented context window: 1.05M, compared with 200K.
The full public result ledger is available for audit without forcing a wide desktop table onto a phone.
Terminal-Bench 2.0
Not directly comparable
OSWorld-Verified
Not directly comparable
Claw-Eval
Not directly comparable
CyberGym
Not directly comparable
Gert Labs
Not directly comparable
OSWorld 2.0
GPT-6 Astra leads this result
JobBench
Not directly comparable
Terminal-Bench 2.1 (Vals)
GPT-6 Astra leads this result
ApprenticeBench
Shared sourceGPT-6 Astra leads this result
MazeBench
GPT-6 Astra leads this result
BrowseComp
Not directly comparable
Terminal-Bench 4.0
Not directly comparable
Terminal-Bench-Science 0.1
Not directly comparable
ExploitGym
Not directly comparable
Agents' Last Exam
Not directly comparable
AutomationBench
Not directly comparable
HLE w/ tools
Not directly comparable
CWE-bench v1
Not directly comparable
Hermes Index
Not directly comparable
Hermes Bench
Not directly comparable
SWE-bench Verified
Not directly comparable
SWE-Rebench
Not directly comparable
React Native Evals
Not directly comparable
Terminal-Bench 2.0
Not directly comparable
Vibe Code Bench
Not directly comparable
cursorBench31
Not directly comparable
FrontierCode 1.1 Main
GPT-6 Astra leads this result
LiveCodeBench (Vals)
Not directly comparable
SWE-bench (Vals)
Not directly comparable
DeepSWE
Not directly comparable
FrontierCode 1.1 Extended
Not directly comparable
FrontierSWE v2
Not directly comparable
PostTrainBench v1.1
Not directly comparable
Bug Hunt Bench
Not directly comparable
ARC-AGI-2
GPT-6 Astra leads this result
ARC-AGI-1
GPT-6 Astra leads this result
STEMBench
Not directly comparable
ARC-AGI-3
Not directly comparable
GeneBench-Pro
Not directly comparable
MRCR v2 256K-512K
Not directly comparable
MRCR v2 512K-1M
Not directly comparable
GraphWalks BFS 256K–1M
Not directly comparable
GPQA
GPT-6 Astra leads this result
SuperGPQA
Not directly comparable
MMLU-Pro
Not directly comparable
HLE
Not directly comparable
GPQA Diamond (Vals)
Not directly comparable
MMLU-Pro (Vals)
Not directly comparable
GPQA-D
Not directly comparable
HealthBench (raw)
Not directly comparable
HealthBench (length-adjusted)
Not directly comparable
HealthBench Professional
Not directly comparable
HealthBench Professional (raw)
Not directly comparable
HealthBench Hard
Not directly comparable
Gray Swan IPI (15 attempts)
Not directly comparable
One weekly email when material rank, price, or benchmark evidence changes make this matchup worth revisiting.
Read a sample issueJoin 5,600+ readers.
Last updated October 17, 2026
