Humansand's Ampersand Coding AI Matches Fable 5.1 at 70% Lower Cost

The multi-agent version of Humansand's Ampersand AI assistant scored 92.2% on the SWE-bench Pro v2 HARD-51 benchmark, matching the results of competitors Fable 5.1 and Gemini 4 Argon. The company said the model achieves this parity without additional coding-specific training and is 70% cheaper to run than Fable 5.1. Ampersand also recorded a 77.9% score on the DeepSWE v1.1 benchmark, placing it ahead of Grok-4.7 and Deepseek-v4-pro.