Skip to content

Compare models

Put up to 4 models beside each other — token prices, context windows, capabilities and provider, from the same catalogue the model pages read.

  1. Devstral 2 2512Mistral AIRemove
  2. Gemini 3.1 Flash Lite Image (Google Vertex AI)GoogleRemove
  3. Kimi K3.1 ProMoonShot AIRemove
devstral-2512 vs gemini-3.1-flash-lite-image-vertex vs kimi-k3.1-pro
AttributeDevstral 2 2512devstral-2512Gemini 3.1 Flash Lite Image (Google Vertex AI)gemini-3.1-flash-lite-image-vertexKimi K3.1 Prokimi-k3.1-pro
Pricing
Input$0.15 / 1M—$2.00 / 1M
Output$0.60 / 1M—$10.00 / 1M
Cache Write (5m)$0.15 / 1MNot applicable$2.00 / 1M
Cache Write (1h)$0.15 / 1MNot applicable$2.00 / 1M
Cache Read$0.15 / 1MNot applicable$2.00 / 1M
Web Search$0 / 1M—$0 / 1M
Request—$0.0336 / request—
Billing—Pay Per Request—
Context
Max context262.1KN/A1M
Max outputN/AN/AN/A
Capabilities
VisionNoYesYes
Function CallingYesNoYes
JSON ModeYesNoYes
StreamingYesNoYes
Catalogue
ProviderMistral AIGoogleMoonShot AI
Categorychatimagechat
Charge typePay As You GoPay Per RequestPay As You Go
Released———
Description
SummaryDevstral 2 is Mistral AI's open-source, agentic coding model — a 123B dense transformer with a 256K context window. It can navigate and modify large codebases, coordinate multi-file changes, manage dependencies, and automatically detect and fix errors. Designed for modernization and large engineering tasks, it’s tunable for specific languages or enterprise code and is released under a modified MIT license.Gemini 3.1 Flash Lite Image served from Google Vertex AI. Billed per image at US$0.0336. Call it through /v1/chat/completions with stream set to false; the image is returned in the message content as a base64 data URI. Maximum output resolution is 1K. A request for a larger size returns a 1K image with the same aspect ratio at the same price. When a size is reduced, the response carries an X-Image-Resolution-Capped header with the requested and delivered size.Kimi K3.1 Pro is Moonshot AI's flagship open-weight multimodal reasoning model, building on Kimi K3 with enhanced post-training for advanced reasoning and autonomous task execution. It features a 1M-token context window and native image and video understanding, enabling complex, long-context multimodal workflows. Optimized for long-horizon coding, 3D creation, financial research, and end-to-end agentic knowledge work, Kimi K3.1 Pro is well suited for autonomous software engineering, multimodal analysis, and complex professional workflows that require sustained reasoning and reliable execution.