Best Model for Pi

Claude Sonnet 5.5 is the best default model for Pi: it scores 61.8% on Terminal-Bench 4.0 at $2/$10 per million tokens, about $5.33 per 100 coding-agent requests on our measured traffic. Pi skips subagents and plan mode and keeps its system prompt short, so one model does all the work; switch to Claude Opus 5.5 (64.9%) mid-session with /model when a task gets hard. DeepSeek V4.1 Flash is the cheapest capable option at about $0.67 per 100 requests at peak rates.

Read the full Pi review → · updated 2026-10-17

Sources: Pi · Terminal-Bench 4.0 leaderboard · Anthropic on Haiku 5.5 · DeepSeek V4.1 Flash release · Cost per request methodology

#1Claude Sonnet 5.5Anthropic · $4.40/1M · 1M ctx

Best overall. 61.8% on Terminal-Bench 4.0 with a 1M-token context, so long Pi sessions run a while before compaction.

Overall 96/100 · Value 59Full breakdown →

#2Claude Opus 5.5Anthropic · $8.80/1M · 1M ctx

Best for long agent runs. #1 on Terminal-Bench 4.0 at 64.9%. Pi lets you switch to it mid-session for the hard part of a task and back again.

Overall 98/100 · Value 51Full breakdown →

#3GPT-6.1 SolOpenAI · $4.40/1M · 1.1M ctx

OpenAI pick. 58.2% on Terminal-Bench 4.0 and 96.1 on Scale's SWE-Bench Pro V2, at the same $2/$10 list price as Sonnet 5.5.

Overall 92/100 · Value 58Full breakdown →

#4Claude Haiku 5.5Anthropic · $0.22/1M · 1M ctx

Best value. Anthropic reports 39.2% on Terminal-Bench 4.0 at $0.10/$0.50 per million tokens, for quick tasks that stay under 100K tokens.

Overall 82/100 · Value 97Full breakdown →

#5DeepSeek V4.1 FlashDeepSeek · $0.22/1M · 1.0M ctx

Budget pick. Open weights at $0.30/$1.20 per million tokens at peak hours and half off-peak; DeepSeek reports 42.3% on Terminal-Bench 4.0 for its October 17 update, up from 31.2%.

Overall 77/100 · Value 93Full breakdown →

#6GLM 5.3Z.AI · $1.50/1M · 1.0M ctx

Open-weight pick. The top open-weight model on Terminal-Bench 4.0 (41.8%), with a 1M-token context.

Overall 83/100 · Value 69Full breakdown →

FAQ

Can you change Pi's model?

Yes. Pi supports 15+ providers and hundreds of models, and you switch mid-session with /model or Ctrl+L. Custom providers, including OpenAI-compatible gateways, are added in models.json or through extensions.

What is the best model for Pi?

Claude Sonnet 5.5 as the default (61.8% on Terminal-Bench 4.0), Claude Opus 5.5 for the hardest tasks, and DeepSeek V4.1 Flash when cost matters most.

Is Claude Sonnet 4.6 still a good default for Pi?

Claude Sonnet 5.5 has replaced it at a lower list price ($2/$10 per million tokens against $3/$15).

Every model on this page runs on Standard Compute

Standard Compute

Current frontier models Claude Fable 5.1 · GPT-6 Astra