Skip to main content
BenchLM

Claude Sonnet 5.5 vs GPT-6 Astra

Updated September 28, 2026. Rank says GPT-6 Astra is ahead. Price, access, and your workload can each overturn that. Public scores include evidence status and uncertainty.

Share or export
Share on XLinkedInSocial cardCSVJSON

Decision reading

GPT-6 Astra has the higher public score estimate, 88.5 versus 80.49, but the 90% score intervals overlap. Treat that as a lead, not a settled winner. 12 results are shared. Category rows resting on Estimated evidence or different benchmark sets are marked directional and do not name a winner.

Model A
Anthropic logo
Claude Sonnet 5.5

Anthropic

80.49/100

Estimated · Public rank #5

90% interval 69.0–92.0

Model B
OpenAI logo
GPT-6 Astra

OpenAI

88.5/100

Supported · Public rank #1

90% interval 84.0–93.0

Shared results
12
Claude Sonnet 5.5 only
35
GPT-6 Astra only
18
Like-for-like categories
2 / 8
Estimated: Claude Sonnet 5.5 · Supported: GPT-6 AstraHow the comparison works

Which one for your work

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.

  • Agentic work

    Tool use, computer use, and multi-step task completion

    GPT-6 Astra

    GPT-6 Astra leads on the public agentic lane, 70.7 to 66.1, with Supported evidence for both models, although the 90% intervals overlap.

    Confidence: limited

  • Long documents

    Prompts that approach the documented context limit

    GPT-6 Astra

    GPT-6 Astra has the larger documented context window.

    Confidence: documented

  • Chat turn cost

    1K fresh input + 500 output tokens

    Claude Sonnet 5.5

    Claude Sonnet 5.5 has the lower estimated token cost for this stated workload. Costs use the listed standard API rates.

    Confidence: listed-rates
Show secondary and unsupported calls

  • Cache-heavy agent loop cost

    200K cached + 20K fresh input + 10K output tokens

    Claude Sonnet 5.5

    Claude Sonnet 5.5 has the lower estimated token cost for this stated workload. Costs use the listed standard API rates.

    Confidence: listed-rates

  • Repository review cost

    50K fresh input + 3K output tokens

    Claude Sonnet 5.5

    Claude Sonnet 5.5 has the lower estimated token cost for this stated workload. Costs use the listed standard API rates.

    Confidence: listed-rates

  • Coding work

    Code generation, repair, and software-engineering tasks

    Not enough matched evidence

    Claude Sonnet 5.5 is scored on Estimated evidence for coding, so the reading is directional rather than like-for-like.

    Confidence: limited

Which one for a specific job

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.

79.6Claude Sonnet 5.574.0GPT-6 Astra

Directional only · BenchAlign v5.7

Claude Sonnet 5.5 scores higher, but at least one score rests on Estimated evidence or a different benchmark set. Directional only, no winner.

Coding scores combine specific tasks and setups. Match the editor, harness, and effort to your workflow.

Open the coding leaderboardShortlist models for this job
Same basis rules as the category table below

What is actually comparable

Shared results can support a head-to-head reading. Results present for only one model describe coverage, not superiority.

2 categories rest on Estimated evidence or different evidence sets. Those rows remain visible for coverage context but do not name a winner.

Shape of the matched evidence

Only shared public evidence is shown. Sparse evidence stays a ruled list rather than being closed into a radar shape.

A shared-evidence shape is not available.

BenchLM does not draw a radar or infer missing axes when the matched evidence is too sparse.

Category results, on a stated basis

Each row shows the public-lane category score for both models: the BenchAlign v5.7 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.

Agentic

Like-for-like
Claude Sonnet 5.5
66.1
Supported · #10/111
GPT-6 Astra
70.7
Supported · #5/111
Basis
BenchAlign v5.7 lane · 11 vs 10 public rows
Reading
GPT-6 Astra leads · intervals overlap

Knowledge

Like-for-like
Claude Sonnet 5.5
80.9
Supported · #6/160
GPT-6 Astra
86.8
Supported · #2/160
Basis
BenchAlign v5.7 lane · 16 vs 7 public rows
Reading
GPT-6 Astra leads · intervals overlap

Coding

Directional only
Claude Sonnet 5.5
79.6
Estimated · #3/136
GPT-6 Astra
74.0
Supported · #4/136
Basis
BenchAlign v5.7 lane · 9 vs 4 public rows
Reading
Directional only

Reasoning

Directional only
Claude Sonnet 5.5
79.2
#7/27
GPT-6 Astra
89.6
#1/27
Basis
Provisional lane · 0 vs 2 weighted rows
Reading
Directional only

Multimodal

Not comparable
Claude Sonnet 5.5
83.3
Unranked · 7 rankable rows
GPT-6 Astra
83.9
Unranked · 3 rankable rows
Basis
Provisional lane · 1 vs 0 weighted rows
Reading
Not comparable

Multilingual

Not comparable
Claude Sonnet 5.5
Not ranked
GPT-6 Astra
Not ranked
Basis
Provisional lane · 0 vs 0 weighted rows
Reading
Not comparable

Instruction following

Not comparable
Claude Sonnet 5.5
Not ranked
GPT-6 Astra
Not ranked
Basis
Provisional lane · 0 vs 0 weighted rows
Reading
Not comparable

Math

Not comparable
Claude Sonnet 5.5
Not ranked
GPT-6 Astra
85.0
Unranked · 1 rankable row
Basis
Provisional lane · 0 vs 1 weighted rows
Reading
Not comparable

Ranks count the models scored in each category’s lane, so the agentic, coding, and knowledge denominators (BenchAlign v5.7) 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.

Supported evidence per lane · bars run 0–100Methodology

What each workload costs

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.

Chat turn

1K fresh input + 500 output tokens

Claude Sonnet 5.5
$0.007
Fits in one request
GPT-6 Astra
$0.035
Fits in one request

Claude Sonnet 5.5 has the lower modeled cost

Costs use the listed standard API rates.

Repository review

50K fresh input + 3K output tokens

Claude Sonnet 5.5
$0.13
Fits in one request
GPT-6 Astra
$0.65
Fits in one request

Claude Sonnet 5.5 has the lower modeled cost

Costs use the listed standard API rates.

Cache-heavy agent loop

200K cached + 20K fresh input + 10K output tokens

Claude Sonnet 5.5
$0.18
Fits in one request
GPT-6 Astra
$0.9
Fits in one request

Claude Sonnet 5.5 has the lower modeled cost

Costs use the listed standard API rates.

Cached input falls back to the list input rate only where a cached rate is unpublished

Specification differences

Sourced differences are shown directly. Missing facts stay explicit instead of being inferred from a model name or family.

Reasoning profile

Claude Sonnet 5.5

Reasoning

GPT-6 Astra

Reasoning

Weight access

Claude Sonnet 5.5

Proprietary

GPT-6 Astra

Proprietary

License

Claude Sonnet 5.5

Proprietary

GPT-6 Astra

Proprietary

Release date

Claude Sonnet 5.5

2026-09-28

GPT-6 Astra

2026-09-03

If you already use one of these models

Deployment change
The models list different providers, so authentication, endpoint behavior, limits, and feature support may change.
Quality signal
GPT-6 Astra has the higher public score estimate, 88.5 versus 80.49, but the 90% score intervals overlap.
Workload cost
Repository review: $0.13 vs $0.65. Cache-heavy agent loop: $0.18 vs $0.9.
Context tradeoff
GPT-6 Astra has the larger documented window (1.05M).
Run the same representative tasks against both endpoints before changing production traffic.

Questions

Which is better, Claude Sonnet 5.5 or GPT-6 Astra?

GPT-6 Astra has the higher public score estimate, 88.5 versus 80.49, but the 90% score intervals overlap. The higher estimate is not a decisive winner because the uncertainty ranges overlap.

Which is better for coding, Claude Sonnet 5.5 or GPT-6 Astra?

Claude Sonnet 5.5 scores higher for coding on the public lane, 79.6 to 74. Claude Sonnet 5.5 is scored on Estimated evidence for coding, so the reading is directional rather than like-for-like. BenchLM does not name a winner for coding from a directional reading; read the shared benchmark rows directly and test the models on the same task set.

Which is better for agentic tasks, Claude Sonnet 5.5 or GPT-6 Astra?

GPT-6 Astra leads the public agentic tasks lane, 70.7 to 66.1, with Supported evidence for both models, although the 90% intervals overlap.

Which costs less, Claude Sonnet 5.5 or GPT-6 Astra?

For the stated presets, chat costs $0.007 on Claude Sonnet 5.5 and $0.035 on GPT-6 Astra; repository review costs $0.13 and $0.65; the cache-heavy agent loop costs $0.18 and $0.9. Costs use the listed standard API rates.

Which has the larger context window, Claude Sonnet 5.5 or GPT-6 Astra?

GPT-6 Astra has the larger documented context window: 1.05M, compared with 1M.

Benchmark evidence

The full public result ledger is available for audit without forcing a wide desktop table onto a phone.

Browse raw public benchmark evidence65 rows

Agentic

  • Terminal-Bench 4.0

    Claude Sonnet 5.570.60%
    Source
    GPT-6 Astra57.90%
    Source

    Claude Sonnet 5.5 leads this result

  • HLE w/ tools

    Claude Sonnet 5.564.5%
    Source
    GPT-6 Astra57.2%
    Source

    Claude Sonnet 5.5 leads this result

  • DRACO

    Claude Sonnet 5.587.0%
    Source
    GPT-6 Astra—

    Not directly comparable

  • Terminal-Bench-Science 0.1

    Claude Sonnet 5.559.9%
    Source
    GPT-6 Astra64.6%
    Source

    GPT-6 Astra leads this result

  • LAB all-pass (Harvey held-out)

    Claude Sonnet 5.510.0%
    Source
    GPT-6 Astra—

    Not directly comparable

  • LAB criterion-pass (Harvey held-out)

    Claude Sonnet 5.593.1%
    Source
    GPT-6 Astra—

    Not directly comparable

  • Toolathlon Verified Pass@3

    Claude Sonnet 5.585.2%
    Source
    GPT-6 Astra—

    Not directly comparable

  • Toolathlon Verified Pass³

    Claude Sonnet 5.568.5%
    Source
    GPT-6 Astra—

    Not directly comparable

  • Toolathlon Verified avg. turns

    Claude Sonnet 5.531.6 turns
    Source
    GPT-6 Astra—

    Not directly comparable

  • AutomationBench (Zapier 1.0.6)

    Claude Sonnet 5.544.7%
    Source
    GPT-6 Astra—

    Not directly comparable

  • Toolathlon-Verified

    Claude Sonnet 5.577.8%
    Source
    GPT-6 Astra—

    Not directly comparable

  • BrowseComp

    Claude Sonnet 5.5—
    GPT-6 Astra91.5%
    Source

    Not directly comparable

  • OSWorld 2.0

    Claude Sonnet 5.5—
    GPT-6 Astra72.6%
    Source

    Not directly comparable

  • ExploitGym

    Claude Sonnet 5.5—
    GPT-6 Astra42.4%
    Source

    Not directly comparable

  • Agents' Last Exam

    Claude Sonnet 5.5—
    GPT-6 Astra59.3%
    Source

    Not directly comparable

  • AutomationBench

    Claude Sonnet 5.5—
    GPT-6 Astra41.4%
    Source

    Not directly comparable

  • Terminal-Bench 2.1 (Vals)

    Claude Sonnet 5.5—
    GPT-6 Astra87.3%
    Source

    Not directly comparable

  • ApprenticeBench

    Claude Sonnet 5.5—
    GPT-6 Astra68%
    Source

    Not directly comparable

Coding

  • FrontierCode 1.1 Main

    Claude Sonnet 5.546.2%
    Source
    GPT-6 Astra53.3%
    Source

    GPT-6 Astra leads this result

  • cursorBench40

    Claude Sonnet 5.555.5%
    Source
    GPT-6 Astra—

    Not directly comparable

  • SWE-bench Pro

    Claude Sonnet 5.581.3%
    Source
    GPT-6 Astra—

    Not directly comparable

  • SWE Multilingual

    Claude Sonnet 5.590.3%
    Source
    GPT-6 Astra—

    Not directly comparable

  • SWE Multimodal

    Claude Sonnet 5.554.3%
    Source
    GPT-6 Astra—

    Not directly comparable

  • DeepSWE

    Claude Sonnet 5.571.0%
    Source
    GPT-6 Astra74.1%
    Source

    GPT-6 Astra leads this result

  • FrontierCode 1.1 Extended

    Claude Sonnet 5.559.1%
    Source
    GPT-6 Astra64.5%
    Source

    GPT-6 Astra leads this result

  • ProgramBench

    Claude Sonnet 5.579.7%
    Source
    GPT-6 Astra—

    Not directly comparable

  • FrontierSWE v2

    Claude Sonnet 5.561.9%
    Source
    GPT-6 Astra65.5%
    Source

    GPT-6 Astra leads this result

Reasoning

  • ARC-AGI-1

    Claude Sonnet 5.5—
    GPT-6 Astra98.50%
    Source

    Not directly comparable

  • ARC-AGI-2

    Claude Sonnet 5.5—
    GPT-6 Astra95%
    Source

    Not directly comparable

  • ARC-AGI-3

    Claude Sonnet 5.5—
    GPT-6 Astra62.7%
    Source

    Not directly comparable

  • GeneBench-Pro

    Claude Sonnet 5.5—
    GPT-6 Astra37.8%
    Source

    Not directly comparable

  • MRCR v2 256K-512K

    Claude Sonnet 5.5—
    GPT-6 Astra100.0%
    Source

    Not directly comparable

  • MRCR v2 512K-1M

    Claude Sonnet 5.5—
    GPT-6 Astra96.3%
    Source

    Not directly comparable

Multimodal

  • Chartography (no tools)

    Claude Sonnet 5.561.6%
    Source
    GPT-6 Astra—

    Not directly comparable

  • Chartography (tools)

    Claude Sonnet 5.590.2%
    Source
    GPT-6 Astra—

    Not directly comparable

  • BenchCAD Vision2Code (no tools)

    Claude Sonnet 5.50.747
    Source
    GPT-6 Astra—

    Not directly comparable

  • BenchCAD Vision2Code (tools)

    Claude Sonnet 5.50.963
    Source
    GPT-6 Astra0.959
    Source

    Claude Sonnet 5.5 leads this result

  • Biomedical image analysis

    Claude Sonnet 5.572.2%
    Source
    GPT-6 Astra—

    Not directly comparable

  • OfficeQA

    Claude Sonnet 5.576.9%
    Source
    GPT-6 Astra—

    Not directly comparable

  • OfficeQA Pro

    Claude Sonnet 5.565.6%
    Source
    GPT-6 Astra—

    Not directly comparable

  • ScreenSpot Pro

    Claude Sonnet 5.5—
    GPT-6 Astra92.7%
    Source

    Not directly comparable

Knowledge

  • HealthBench (raw)

    Claude Sonnet 5.569.4%
    Source
    GPT-6 Astra56.9%
    Source

    Claude Sonnet 5.5 leads this result

  • HealthBench (length-adjusted)

    Claude Sonnet 5.565.4%
    Source
    GPT-6 Astra58.3%
    Source

    Claude Sonnet 5.5 leads this result

  • HealthBench Professional (raw)

    Claude Sonnet 5.577.1%
    Source
    GPT-6 Astra68.2%
    Source

    Claude Sonnet 5.5 leads this result

  • HealthBench Professional

    Claude Sonnet 5.569.2%
    Source
    GPT-6 Astra64.7%
    Source

    Claude Sonnet 5.5 leads this result

  • BioMysteryBench (human-solvable)

    Claude Sonnet 5.589.2%
    Source
    GPT-6 Astra—

    Not directly comparable

  • BioMysteryBench (human-difficult)

    Claude Sonnet 5.544.7%
    Source
    GPT-6 Astra—

    Not directly comparable

  • SpatialBench Verified

    Claude Sonnet 5.572.5%
    Source
    GPT-6 Astra—

    Not directly comparable

  • SingleCellBench

    Claude Sonnet 5.559.1%
    Source
    GPT-6 Astra—

    Not directly comparable

  • Protein Design

    Claude Sonnet 5.551.0%
    Source
    GPT-6 Astra—

    Not directly comparable

  • Morphology-to-molecule matching

    Claude Sonnet 5.525.0%
    Source
    GPT-6 Astra—

    Not directly comparable

  • Medicinal chemistry

    Claude Sonnet 5.565.3%
    Source
    GPT-6 Astra—

    Not directly comparable

  • Protein Design library ranking

    Claude Sonnet 5.554.8%
    Source
    GPT-6 Astra—

    Not directly comparable

  • De novo protein-binder design

    Claude Sonnet 5.582.3%
    Source
    GPT-6 Astra—

    Not directly comparable

  • Protocols (troubleshooting)

    Claude Sonnet 5.567.3%
    Source
    GPT-6 Astra—

    Not directly comparable

  • Protocols (understanding)

    Claude Sonnet 5.566.6%
    Source
    GPT-6 Astra—

    Not directly comparable

  • HLE w/o tools

    Claude Sonnet 5.556.9%
    Source
    GPT-6 Astra—

    Not directly comparable

  • GPQA

    Claude Sonnet 5.5—
    GPT-6 Astra96%
    Source

    Not directly comparable

  • GPQA-D

    Claude Sonnet 5.5—
    GPT-6 Astra96.0%
    Source

    Not directly comparable

  • HealthBench Hard

    Claude Sonnet 5.5—
    GPT-6 Astra36.6%
    Source

    Not directly comparable

Multilingual

  • GMMLU

    Claude Sonnet 5.592.1%
    Source
    GPT-6 Astra—

    Not directly comparable

  • MILU

    Claude Sonnet 5.591.6%
    Source
    GPT-6 Astra—

    Not directly comparable

Math

  • ArXivMath Aug. 2026 (no tools)

    Claude Sonnet 5.586.8%
    Source
    GPT-6 Astra—

    Not directly comparable

  • ArXivMath Aug. 2026 (tools)

    Claude Sonnet 5.595.2%
    Source
    GPT-6 Astra—

    Not directly comparable

  • FrontierMath v2 (Tier 4)

    Claude Sonnet 5.5—
    GPT-6 Astra97.600%
    Source

    Not directly comparable

65 public results · 12 shared

Related comparisons

Compare API pricingRead the methodologyOpen the model selector

Watch Claude Sonnet 5.5 vs GPT-6 Astra

One weekly email when material rank, price, or benchmark evidence changes make this matchup worth revisiting.

Read a sample issue

Join 2,000+ readers.

Last updated September 28, 2026