The Known Good Updated 2 Sep 2026 Subscribe
Anthropic: Claude Sonnet 4.6 vs OpenAI: GPT-5.5 ·Anthropic vs OpenAI

Anthropic: Claude Sonnet 4.6 vs OpenAI: GPT-5.5 — head-to-head comparison

Anthropic: Claude Sonnet 4.6 profile OpenAI: GPT-5.5 profile All comparisons
Anthropic: Claude Sonnet 4.6 — Anthropic
Quality
#18 of 65
81based on 4 of 7 evaluations
Known Good Index
Speed
#205 of 339
40.9
Output tokens per second
Price
#391 of 439
$6.00
Blended per 1M tokens
In $3.00 · Out $15.00
Cache hit price
#277 of 322
$0.30
per 1M tokens (−90%)
Context
#114 of 473
1.00M
tokens (~1,500 pages)
OpenAI: GPT-5.5 — OpenAI
Quality
#5 of 65
93based on 3 of 7 evaluations
Known Good Index
Speed
#65 of 339
85.0
Output tokens per second
Price
#412 of 439
$11.25
Blended per 1M tokens
In $5.00 · Out $30.00
Cache hit price
#295 of 322
$0.50
per 1M tokens (−90%)
Context
#30 of 473
1.05M
tokens (~1,600 pages)
Summary

On the Known Good Index, Anthropic: Claude Sonnet 4.6 scores 81 and OpenAI: GPT-5.5 scores 93 — a 12-point lead for OpenAI: GPT-5.5. They share 3 published evaluations, with Anthropic: Claude Sonnet 4.6 ahead on 1 and OpenAI: GPT-5.5 ahead on 2.

Blended pricing is $6.00 per 1M tokens for Anthropic: Claude Sonnet 4.6 and $11.25 for OpenAI: GPT-5.5 — Anthropic: Claude Sonnet 4.6 is about 1.9× cheaper.

Throughput is 41 tokens per second for Anthropic: Claude Sonnet 4.6 and 85 for OpenAI: GPT-5.5, with time to first token at 1.6s and 4.9s. Anthropic: Claude Sonnet 4.6 accepts 1.00M tokens of context against 1.05M.

Anthropic: Claude Sonnet 4.6 is proprietary; OpenAI: GPT-5.5 is proprietary. 6 tracked hosts serve one or both, and 2 serve both.

Generated from tracked data on 3 Sep 2026 · how these comparisons are computed →

At a glance
MetricAnthropic: Claude Sonnet 4.6OpenAI: GPT-5.5
Known Good Index 81 based on 4 of 7 evaluations 93 based on 3 of 7 evaluations
Blended $/1M $6.00 $11.25
Output tok/s 41 85
Context window 1.00M 1.05M
Providers 5 3
Jump to section

Quality

Known Good Index and the capability indices, side by side

Index scores

Each index is min-max normalised across tracked models and scaled 0–100 from published evaluations. We do not run these evaluations.

MetricAnthropic: Claude Sonnet 4.6OpenAI: GPT-5.5
Known Good Index 81 93
Coding Index 73
Openness Index 5 5
The Known Good

Quality Breakdown

Every evaluation both models report a published score for

Evaluations both models report

Bar = Anthropic: Claude Sonnet 4.6 · marker = OpenAI: GPT-5.5 · scores ingested from Epoch AI (CC-BY 4.0)

GPQA Diamond87 / 91
Mock AIME 2024–2586 / 84
Terminal-Bench53 / 85

Right-hand figures read Anthropic: Claude Sonnet 4.6 / OpenAI: GPT-5.5. Anthropic: Claude Sonnet 4.6 leads on 1 of 3, OpenAI: GPT-5.5 on 2.

The Known Good

Arena Elo

Human preference rating from blind pairwise votes

Arena ratings

Elo from blind pairwise votes, ingested from Arena. Higher is better.

MetricAnthropic: Claude Sonnet 4.6OpenAI: GPT-5.5
Text Arena 1,458 1,467
Text Arena (style controlled) 1,472 1,477
The Known Good

Price & Cost

Published API pricing by token type

Published pricing

USD per 1M tokens. Blended is a 3:1 input:output mix. Live from OpenRouter, cross-checked against models.dev. Lower is better.

MetricAnthropic: Claude Sonnet 4.6OpenAI: GPT-5.5
Input $/1M $3.00 $5.00
Output $/1M $15.00 $30.00
Blended $/1M $6.00 $11.25
Cache read $/1M $0.300 $0.500
The Known Good
We publish blended price rather than cost-per-task. Cost-per-task requires token counts from an evaluation suite we do not run. Blended price is computed from published rates and is fully reproducible — see the pricing methodology.

Context Window

Maximum input accepted and maximum output emitted

Context and output limits

Provider-declared limits, from metadata via models.dev and OpenRouter. Higher is better.

MetricAnthropic: Claude Sonnet 4.6OpenAI: GPT-5.5
Context window 1.00M 1.05M
Max output tokens 128k 128k
The Known Good

Speed & Latency

Mean throughput and time to first token across hosting providers, over successful measurements from the last 48 hours

Throughput and latency

Means across all tracked providers over successful measurements from the last 48 hours, derived from OpenRouter endpoint telemetry.

MetricAnthropic: Claude Sonnet 4.6OpenAI: GPT-5.5
Output tokens/s 41 85
Time to first token 1.6s 4.9s
The Known Good

Providers

Which hosts serve each model, and at what price and speed

6 hosts · 2 serve both
Full provider benchmarks →
Provider Anthropic: Claude Sonnet 4.6 blendedAnthropic: Claude Sonnet 4.6 tok/s OpenAI: GPT-5.5 blendedOpenAI: GPT-5.5 tok/s
Amazon Bedrock · first party $6.00 49 $12.38 180
Claude Platform on AWS $6.00 41
Google · first party $6.00 40
Anthropic · first party $6.00 31
Azure · first party $6.00 30 $11.25 54
OpenAI · first party $5.62 70

Per-host pricing and telemetry from OpenRouter. A dash means that host does not serve that model.

The Known Good
Sources for this page
Pricing & providers — OpenRouter Metadata — models.dev Benchmarks — Epoch AI (CC-BY 4.0)Elo — Arena Full attribution →