Claude Sonnet 5.5 vs GPT-6.1 Sol

$0.88 / task

Wins 3 of 17 benchmarks

$0.21 / task

Wins 12 of 17 benchmarks

Different efforts: high · medium · Compare all at high

Only shared

AutomationBench-AA AutomationBench-AA: Agents orchestrate SaaS workflows through APIs; objectives receive credit only when the task has no guardrail violation

Sonnet 5.559.4

GPT-6.1 Sol62.6

GDPval-AA (work tasks) 2.1 GDPval-AA (work tasks) 2.1: Real work tasks from many professions (reports, spreadsheets, analyses), graded against expert work

Sonnet 5.552.5

GPT-6.1 Sol47.5

AA Analyst Agent AA Analyst Agent: Data-analyst tasks: the model explores data with tools and reports findings

Sonnet 5.5 · max57.5

GPT-6.1 Sol · max50

Terminal-Bench 4.0 Terminal-Bench 4.0 · source setting 2: Agents use a terminal to complete executable coding, system administration and research tasks in the 4.0 suite

Sonnet 5.543.9

GPT-6.1 Sol48

Terminal-Bench Science 0.1 Terminal-Bench Science 0.1: Agents complete expert-authored research workflows across five scientific domains in a terminal

Sonnet 5.534.3

GPT-6.1 Sol · max58.1

AA-Briefcase 1.1 · Elo AA-Briefcase 1.1 · Elo: Agents produce professional deliverables; blind pairwise comparisons produce the source Elo rating

Sonnet 5.51,635.8

GPT-6.1 Sol1,361.2

AA-LCR (long context) 1.1 AA-LCR (long context) 1.1: Questions that need reading and connecting facts across very long documents (around 100K tokens)

Sonnet 5.578

GPT-6.1 Sol83.3

DeepSWE 1.1 DeepSWE 1.1: Agents change repository code to resolve software engineering tasks under the stated DeepSWE edition

Sonnet 5.5 · max70.5

GPT-6.1 Sol · max72.3

FrontierCode 1.1 FrontierCode 1.1: Agents implement changes in real repositories, graded against maintainer-defined requirements

Sonnet 5.544.3

GPT-6.1 Sol · max49.1

Harvey LAB-AA 1.1 Harvey LAB-AA 1.1: Legal agents work from case documents; every rubric criterion must pass and material hallucinations receive no credit

Sonnet 5.5 · max2.8

GPT-6.1 Sol · max6.9

SWE-bench Pro (v2 full) SWE-bench Pro (v2 full): Agents resolve repository issues under the network-locked v2 protocol; full and 51-task Hard results are separate

Sonnet 5.5 · xhigh95.8

GPT-6.1 Sol · xhigh96.1

Furniture Assembly Furniture Assembly: Models inspect furniture assembly photos, identify incorrect builds and locate the mistaken step

Sonnet 5.5 · max75

GPT-6.1 Sol · max80

SpireClimb SpireClimb: Agents play Slay the Spire under the named game and difficulty setting

Sonnet 5.5 · max1.3

GPT-6.1 Sol · max24.7

SpireClimb StS1 A20 Heart SpireClimb StS1 A20 Heart: Agents play Slay the Spire under the named game and difficulty setting

Sonnet 5.5 · max0

GPT-6.1 Sol · max20

SpireClimb StS2 A10 SpireClimb StS2 A10: Agents play Slay the Spire under the named game and difficulty setting

Sonnet 5.5 · max2.7

GPT-6.1 Sol · max29.3

SWE-bench Pro (v2 hard) SWE-bench Pro (v2 hard): Agents resolve repository issues under the network-locked v2 protocol; full and 51-task Hard results are separate

Sonnet 5.5 · xhigh70.6

GPT-6.1 Sol—

EBR Bench EBR Bench: Agents learn to play Earthborne Rangers across repeated playthroughs

Sonnet 5.5—

GPT-6.1 Sol · max54.3

Benchmark	Sonnet 5.5	GPT-6.1 Sol
AutomationBench-AA	59.4	62.6
GDPval-AA (work tasks) 2.1	52.5	47.5
AA Analyst Agent	57.5 · max	50 · max
Terminal-Bench 4.0	43.9	48
Terminal-Bench Science 0.1	34.3	58.1 · max
AA-Briefcase 1.1 · Elo	1,635.8	1,361.2
AA-LCR (long context) 1.1	78	83.3
DeepSWE 1.1	70.5 · max	72.3 · max
FrontierCode 1.1	44.3	49.1 · max
Harvey LAB-AA 1.1	2.8 · max	6.9 · max
SWE-bench Pro (v2 full)	95.8 · xhigh	96.1 · xhigh
Furniture Assembly	75 · max	80 · max
SpireClimb	1.3 · max	24.7 · max
SpireClimb StS1 A20 Heart	0 · max	20 · max
SpireClimb StS2 A10	2.7 · max	29.3 · max
SWE-bench Pro (v2 hard)	70.6 · xhigh	—
EBR Bench	—	54.3 · max

Specs

Spec	Sonnet 5.5	GPT-6.1 Sol
Input, per 1M	$2	$2
Output, per 1M	$10	$10
Cost per task?	$0.88	$0.21
Speed, tok/s	106	45
Context	1M	1.05M
Weights	Closed	Closed
Released	Sep 2026	Sep 2026
See images	✓	✓
Read PDFs	✓	Partial
Hear audio	—	—
Watch video	—	—
Call tools	✓	✓
Force a tool call	—	Partial
Return strict JSON	✓	✓
Search the web	✓	✓
Run code	✓	✓
Use a computer	✓	✓

What breaks when you switch

Sonnet 5.5 → GPT-6.1 Sol

Remove unsupported parameters›

With reasoning enabled, remove temperature, top_p and top_logprobs, plus Chat Completions logprobs or Responses message.output_text.logprobs includes.

Responses API for tool calling · Chat Completions›

Use Responses for tools; Chat Completions supports this model without tool calling.

No none or minimal reasoning›

The none and minimal reasoning efforts are unsupported; choose low, medium, high, xhigh or max.

More than 272K input tokens›

Prompts above 272K input tokens charge 2× input and cache rates and 1.5× output rates for the entire request.

Cache writes billed at 1.25x›

Cache writes cost 1.25× the uncached input token rate.

Ultrafast prices 6x Standard›

Ultrafast mode costs 6× Standard token prices.

GPT-6.1 Sol → Sonnet 5.5

Context: 1.05M → 1M

No forced tool use›

Forced tool_choice any or tool returns HTTP 400; remove the forced choice.

Replace disabled thinking›

Disabled thinking returns HTTP 400; use between_tools for the lowest thinking setting.

No manual thinking budgets›

Manual thinking budgets and extra between_tools fields return HTTP 400; remove manual budgets and send between_tools without display, budget_tokens or block_binding.

Remove sampling parameters›

Non-default temperature, top_p or top_k returns HTTP 400; remove these parameters.

No assistant prefill›

A prefilled final assistant turn returns HTTP 400; remove that prefill.

Old computer tool rejected · Claude API / Google Cloud›

The computer_20251124 tool returns HTTP 400 on the Claude API and Google Cloud; use computer_toolset_20260801 there.