Warm plans, clear limits

Image models bill against monthly image credits. Lower-cost 1K image models sit in the 1-credit tier; higher-resolution or premium-quality settings step up to protect monthly credit pools.

Provider	Model	Reference images	Credits per image
ChatGPT	ChatGPT Image Latest	Up to 8 references	1–6
ChatGPT	GPT Image 2	Up to 8 references	1–6
ChatGPT	GPT Image 1.5	Up to 8 references	1–6
Gemini	Nano Banana / Gemini 2.5 Flash Image	Up to 3 references	2
Gemini	Nano Banana Pro / Gemini 3 Pro Image	Up to 8 references	3–6
Gemini	Nano Banana 2 / Gemini 3.1 Flash Image	Up to 8 references	2–4
Gemini	Imagen 4 Fast	Up to 3 references	1
Gemini	Imagen 4	Up to 3 references	1
Gemini	Imagen 4 Ultra	Up to 3 references	2
Grok	Grok Imagine	Up to 3 references	1
Grok	Grok Imagine Image (Quality)	Up to 3 references	1–2
Black Forest Labs	FLUX.1 Kontext Pro	Up to 8 references	1
Black Forest Labs	FLUX.1 Kontext Max	Up to 8 references	2
Seedream	Seedream 4.5	Up to 10 references	1
Krea	Krea 2 Turbo	Text-to-image	1

All 72 models, each with the provider's own reference pricing so you can compare honestly. Your plan bills in flat request units — the full technical breakdown is below.

OpenAI Pricing, for transparency

All prices shown are per million tokens unless otherwise noted. Serving gpt-4o profitably is challenging given OpenAI's pricing structure. We use summarization sparingly to enable longer conversations, but we will never route you to a different model disingenuously. With us, you always know which model you're using.

Grok Pricing, for transparency

All prices shown are per million tokens. Grok models offer competitive pricing with fast response times and specialized capabilities for reasoning and code generation. For Grok 4.3 and current Grok 4.20, API pricing is $1.25 input, $0.20 cached input, and $2.50 output per 1M tokens; higher context pricing applies above 200K context.

Anthropic Pricing, for transparency

All prices shown are per million tokens. We pass through Anthropic's pricing directly to you. Claude models offer excellent reasoning capabilities and are available to Studio and Aurora tier users.

Gemini Pricing, for transparency

All prices shown are per million tokens unless otherwise noted. Gemini models cover fast base-tier options, long-context Pro models, and newer Flash models with search grounding. Gemini 3.5 Flash is premium and uses 2 premium requests per send before length multipliers.

* Gemini 2.5 Pro pricing varies by prompt size: $1.25/$2.50 input and $10.00/$15.00 output for prompts ≤200k tokens / >200k tokens respectively.
* Gemini 3.1 Pro Preview (1M/64k context, Jan 2025): $2/$4 input and $12/$18 output (including thinking tokens) for prompts ≤200k / >200k tokens respectively.
** Gemini 3.1 Flash-Lite: $0.25/1M for text/image/video input, $0.50/1M for audio; output (including thinking) $1.50/1M. Base model.
*** Gemini 3.5 Flash: $1.50 input and $9.00 output per 1M tokens. Premium model; uses 2 premium requests per send before length multipliers. Search grounding may add Google tool charges after the shared free allowance.

DeepSeek Pricing, for transparency

All prices shown are per million tokens. We use Fireworks for DeepSeek. DeepSeek models via Fireworks offer strong reasoning and agent performance. DeepSeek v3.x and DeepSeek V4 Flash are base models available on all tiers; DeepSeek V4 Pro is Premium. Does not support web search or image input.

Cached input: deepseek-v4-flash $0.03/1M, deepseek-v4-pro $0.14/1M. DeepSeek V4 Flash is a base model with 1M context and function calling; DeepSeek V4 Pro remains Premium with 1M context and function calling.

Qwen Pricing (via Fireworks and Vercel AI Gateway)

Per million tokens. Qwen 3.6 Plus is available through Fireworks serverless on just4o.chat as a base model on all tiers; it uses 4 base requests per send before length multipliers. Qwen 3.7 Plus is served through the Vercel AI Gateway as a base model and uses 2 base requests per send. Both support function calling and image input, but not web search.

Cached input: qwen3p6-plus $0.10/1M. Fireworks also offers dedicated deployments for this model outside the shared serverless route.

GPT OSS Pricing (OpenAI open-weight via Fireworks and Cerebras)

Per million tokens. OpenAI gpt-oss series is available via Fireworks and Cerebras. gpt-oss-120b and gpt-oss-20b are base Fireworks models; gpt-oss-120b-fast is a Premium OSS Cerebras chat-completions variant for faster replies. It uses 1 premium request per send. Does not support web search or image input. gpt-oss-20b does not support function calling.

Kimi Pricing (Moonshot AI via Fireworks and Vercel AI Gateway)

Per million tokens. Moonshot AI Kimi series. K2.5 and K2.6 are served via Fireworks; K2.7 Code and K2.7 Code Highspeed are served through the Vercel AI Gateway. Premium models for reasoning-heavy and multimodal agentic work. Kimi K2.6 and K2.7 Code use 2 premium requests per send before length multipliers; K2.7 Code Highspeed uses 3. Does not support web search. K2.5, K2.6, K2.7 Code, and K2.7 Code Highspeed support image input.

Cached input: kimi-k2p5 $0.10/1M, kimi-k2p6 $0.16/1M.

MiniMax Pricing (via Fireworks and Vercel AI Gateway)

Per million tokens. MiniMax M2.5 and M2.7 are served via Fireworks; M3 is served through the Vercel AI Gateway. Strong coding, agentic tool use, and office work. Base models available on all tiers. Does not support web search.

Cached input: minimax-m2p5 $0.03/1M, minimax-m2p7 $0.06/1M.

Z.ai GLM Pricing (via Fireworks, Cerebras, and Vercel AI Gateway)

Per million tokens. glm-4p7, glm-5, and glm-5p1 are Fireworks models; glm-5p1 uses 2 premium requests per send before length multipliers. glm-5p2 is served through the Vercel AI Gateway and uses 2 premium requests per send. glm-4p7-fast is a Premium OSS Cerebras chat-completions variant that also uses 2 premium requests per send. GLM models do not support web search or image input.

Cached input: glm-4p7 $0.30/1M, glm-5 $0.20/1M, glm-5p1 $0.26/1M, glm-5p2 $0.26/1M. Cerebras currently lists glm-4p7-fast as a preview model.

NVIDIA Nemotron Pricing (via Vercel AI Gateway)

Per million tokens. NVIDIA Nemotron 3 Ultra (550B-A55B) is an open mixture-of-experts model served through the Vercel AI Gateway as a base model on all tiers. It uses 2 base requests per send before length multipliers and supports function calling. Does not support web search or image input.

Mistral Pricing (via Vercel AI Gateway)

Per million tokens. Mistral AI models are served through the Vercel AI Gateway. Mistral Large 3, Devstral 2, Devstral Small 2, and Ministral 14B are base models on all tiers; Mistral Medium 3.5 and Magistral Medium (a native reasoning model) are Premium. Mistral Medium 3.5, Magistral Medium, and Ministral 14B support image input. Mistral applies zero data retention to gateway traffic. These models do not support web search.

Echo (just4o.chat)

Native just4o.chat models. There is no separate per-token API pass-through; usage is covered by your plan limits and request-weighting rules like other chat models.