Proven Weavecuratedchat at scalecost-sensitivespeed

Support Bot at Scale

Hundreds of thousands of conversations a month. Latency and unit economics dominate; a fast, cheap tier handles the floor while classification routes edge cases up.

by WeavePrism · May 23, 2026

Chat

250M in · 100M out$55.0

Classify

100M in · 6M out$7.40

$62.4per month

−$4,338/mo (99%)

Mistral Small 3.2

Mistral

$55.0

$0.1/M in$0.3/M out128k ctx

GPT-5 nano

OpenAI

$7.40

$0.05/M in$0.4/M out400k ctx

React Flow

$62.4/mo woven−$4,338/mo vs Claude Opus 4.5

Assignments

ChatMistral Small 3.2$55.0Q82

Mistral Small 3.2 wins on cost efficiency (99) and speed (100) at $55.0/mo for this volume. Worth testing: GPT-5 nano saves $2.50/mo for −2 quality.

ClassifyGPT-5 nano$7.40Q82

GPT-5 nano wins on cost efficiency (100) and speed (100) at $7.40/mo for this volume.

Trade-off shape

Recomputed live from catalog v2026.01 list pricing and heuristic quality scores.