DeepSWE

All models run on mini-swe-agent harness. Refer to DeepSWE for more details.

1. 1Claude Opus 5[max]74%±4%
2. 2GPT-5.6 Sol[max]73%±5%
3. 3Claude Fable 5[max]70%±4%
4. 4GPT-5.5[xhigh]67%±6%
5. 5Claude Opus 4.8[max]59%±2%
6. 6Grok 4.5[high]53%±5%

Pricing

Per-million-token pricing and prompt-cache rates for every model behind the Revolt endpoint.

OpenAIClaudeSpaceX AI

GPT-5.6 Sol

Context

1.05M

Max output

128K

Input type

file, text, image

Output type

text

ZDR

Yes

Input

$5

Output

$30

Cache read

$0.50

Cache write

$6.25

GPT-5.6 Terra

Context

1.05M

Max output

128K

Input type

file, text, image

Output type

text

ZDR

Yes

Input

$2

Output

$12

Cache read

$0.20

Cache write

$2.50

GPT-5.5

Context

1.05M

Max output

128K

Input type

file, text, image

Output type

text

ZDR

Yes

Input

$5

Output

$30

Cache read

$0.50

Cache write

Free