What are the strengths of this model?

Same $2/$10 as Sonnet 5 but 30% faster and up to 30% cheaper per task Leads Terminal-Bench 4.0 at 70.6%, ahead of Opus 5.5 First Sonnet to clear Pokemon Red from screenshots Best-value Claude for high-volume coding and vision

What are the weaknesses of this model?

Max tier costs ~$7.60 per query, can exceed Opus at full effort Incremental refresh, not a frontier-capability leap Intelligence Index 64 behind Opus 5.5 and Astra Context window not officially disclosed

What are the best use cases?

High-volume CI and code-review bots GUI/vision agents from screenshot reasoning Drop-in upgrade for Sonnet 5 workloads

AnthropicProprietary

Claude Sonnet 5.5

Name: Claude Sonnet 5.5
Author: Anthropic

Compare this model

Claude Sonnet 5.5 is Anthropic's mid-tier model refresh, shipping at the same $2/$10 per-million pricing as Sonnet 5 but 30% faster and up to 30% cheaper per task. It leads Terminal-Bench 4.0 at 70.6%, ahead of its larger sibling Opus 5.5, making it the default choice for high-volume coding and vision agent work.

Share:X Hatena LINE

Parameters

TBA

Context Window

TBA

License

Proprietary

Release Date

2026-09-28

Benchmark Performance

AA Intelligence Index

64.0

LMArena Elo

—

HLE

—

ARC-AGI-2

—

SWE-bench Verified

—

GPQA Diamond

—

MMLU-Pro

—

LiveCodeBench

—

AIME 2025

—

MATH-500

—

API Pricing

Input Price (per 1M tokens)

Output Price (per 1M tokens)

$10

Billing Mode: per token

Strengths

•Same $2/$10 as Sonnet 5 but 30% faster and up to 30% cheaper per task
•Leads Terminal-Bench 4.0 at 70.6%, ahead of Opus 5.5
•First Sonnet to clear Pokemon Red from screenshots
•Best-value Claude for high-volume coding and vision

Weaknesses

•Max tier costs ~$7.60 per query, can exceed Opus at full effort
•Incremental refresh, not a frontier-capability leap
•Intelligence Index 64 behind Opus 5.5 and Astra
•Context window not officially disclosed

Use Cases

•High-volume CI and code-review bots
•GUI/vision agents from screenshot reasoning
•Drop-in upgrade for Sonnet 5 workloads

Deep Analysis

Artificial Analysis Intelligence Index

64 (#2)

second overall on the AA composite

Terminal-Bench 4.0

70.6%

ahead of Opus 5.5's 66.4%

Pricing (input / output)

$2 / $10 per M

identical to Sonnet 5

Speed

30% faster than Sonnet 5

same $2/$10 price

Chartography

15.6% to 61.6%

large jump on the chart-reasoning eval

Strengths

・Same $2/$10 as Sonnet 5 but 30% faster and up to 30% cheaper per task
・Tops Terminal-Bench 4.0 at 70.6%, ahead of Opus 5.5
・First Sonnet to clear Pokemon Red from screenshots alone
・Strong mid-tier value for high-volume agent and vision work

Weaknesses

・Max tier runs ~$7.60 per query, can cost more than Opus at full effort
・Not a frontier-capability leap; it is a Sonnet-class refresh
・Intelligence Index (64) still well behind Opus 5.5 and Astra
・Context window not officially disclosed (TBA)

Competitor Comparison

Model	Arena	SWE	GPQA	Price
Claude Opus 5.5	58	n/a	n/a	$10/$50 per M
GPT-6.1 Sol	52	n/a	n/a	$2/$10 per M
Claude Sonnet 5.5	64	TB 70.6%	n/a	$2/$10 per M

Overview

Claude Sonnet 5.5 is Anthropic's Sep 28, 2026 mid-tier refresh: the same $2/$10 pricing as Sonnet 5 but 30% faster and up to 30% cheaper per task, with a Terminal-Bench 4.0 score that beats its bigger sibling Opus 5.5.

Benchmarks & Performance

On Artificial Analysis' Intelligence Index Sonnet 5.5 scores 64 (#2). It posts 70.6% on Terminal-Bench 4.0, ahead of Opus 5.5's 66.4%, and Chartography rose from 15.6% to 61.6%. It is the first Sonnet to clear Pokemon Red using only screenshots. Max tier runs ~$7.60 per query.

Detailed Comparison

Against Sonnet 5 it is strictly better at equal price. Against Opus 5.5 it wins on cost and on Terminal-Bench but trails on the composite index. Against GPT-6.1 Sol it is six index points ahead but pricier per token. It squeezes the mid-tier from both sides.

Community Feedback

* Praise: developers call it the best-value Claude for high-volume coding and vision. * Criticism: Max-tier per-query cost draws complaints; some note it is an incremental refresh, not a capability leap. * Practical: teams route routine work to Sonnet 5.5 and reserve Opus 5.5 for the hardest reasoning.

Use Cases

### Where Sonnet 5.5 earns its keep **1. High-volume coding** - 30% faster at lower cost fits CI and review bots. **2. Vision agents** - Pokemon-Red-clearing screenshot reasoning suits GUI automation. **3. Cost-bound production** - same $2/$10 as Sonnet 5 means a drop-in upgrade.

Latest News

**Released (2026-09-28).** Anthropic shipped Claude Sonnet 5.5 at $2/$10, 30% faster and up to 30% cheaper per task than Sonnet 5, with Terminal-Bench 4.0 at 70.6% (Anthropic).

Sources

Analysis generated: 2026-10-01