Pricing

Credits per model

Every model your assistants can use on StellarFirm, and the credits it takes to read and to write a million tokens. Bigger models use more credits per word, smaller ones fewer. Your own model keys never use credits.

Back to pricing

Credits per million tokens
Model	Maker	Reading	From cache	Writing	Included in
Claude Fable 5.1Anthropic	Anthropic	1,429	35.73	7,145	Trial, Plus and Ultra
Claude Haiku 5.5Anthropic	Anthropic	14.29	1.43	71.45	Trial, Plus and Ultra
Claude Opus 4.8Anthropic	Anthropic	715	71.45	3,573	Trial, Plus and Ultra
Claude Opus 5.5Anthropic	Anthropic	572	28.58	2,858	Trial, Plus and Ultra
Claude Sonnet 5.5Anthropic	Anthropic	286	28.58	1,429	Trial, Plus and Ultra
GPT-5.6 LunaOpenAI	OpenAI	31.44	3.14	189	Trial, Plus and Ultra
GPT-5.6 TerraOpenAI	OpenAI	314	31.44	1,886	Trial, Plus and Ultra
GPT-6 AstraOpenAI	OpenAI	1,572	157	7,860	Trial, Plus and Ultra
GPT-6 LunaOpenAI	OpenAI	15.72	1.57	78.6	Trial, Plus and Ultra
GPT-6 SolOpenAI	OpenAI	314	31.44	1,572	Trial, Plus and Ultra
GPT-6.1 SolOpenAI	OpenAI	314	15.72	1,572	Trial, Plus and Ultra
GPT-6.1 TerraOpenAI	OpenAI	157	7.86	786	Trial, Plus and Ultra
Gemini 3.1 ProGoogle	Google	286	28.58	1,715	Trial, Plus and Ultra
Gemini 3.5 Flash-LiteGoogle	Google	42.87	4.29	357	Trial, Plus and Ultra
Gemini 3.8 FlashGoogle	Google	107	10.72	536	Trial, Plus and Ultra
Gemini 4 ArgonGoogle	Google	286	14.29	1,429	Trial, Plus and Ultra
Grok 4.7xAI	xAI	286	71.45	857	Trial, Plus and Ultra
Grok BuildxAI	xAI	143	28.58	286	Trial, Plus and Ultra
CodestralMistral AI	Mistral AI	59.99	6	180	Every plan
Ministral 3 8BGreenMistral AI	Mistral AI	30	3	30	Every plan
Ministral 3 14BGreenMistral AI	Mistral AI	40	4	40	Every plan
Mistral Large 3Mistral AI	Mistral AI	99.99	10	300	Every plan
Mistral Large 4Mistral AI	Mistral AI	140	14	420	Every plan
Mistral Medium 3.5Mistral AI	Mistral AI	300	30	1,500	Every plan
Mistral Small 4GreenMistral AI	Mistral AI	30	3	120	Every plan
Llama 4 ScoutMeta	Meta	18.18	Same as reading	54.54	Every plan
Muse Glimmer 30BMeta	Meta	54.54	Same as reading	218	Every plan
Muse Spark 1.3Meta	Meta	227	27.27	773	Every plan
Muse WonderMeta	Meta	727	Same as reading	3,636	Every plan
DeepSeek V4 FlashDeepSeek	DeepSeek	16.36	Same as reading	32.72	Every plan
DeepSeek V4 ProDeepSeek	DeepSeek	236	Same as reading	473	Every plan
DeepSeek V4.1 FlashGreenDeepSeek	DeepSeek	36.36	Same as reading	109	Every plan
DeepSeek V4.1 Flash 1017GreenDeepSeek	DeepSeek	59.99	3.82	240	Every plan
Qwen 3.7 PlusAlibaba Cloud	Alibaba Cloud	72.72	Same as reading	291	Every plan
Qwen 3.8 2.4TAlibaba Cloud	Alibaba Cloud	364	Same as reading	1,091	Every plan
Qwen 3.8 FlashAlibaba Cloud	Alibaba Cloud	27.27	Same as reading	85.45	Every plan
Qwen 3.8 MaxAlibaba Cloud	Alibaba Cloud	364	45.45	1,091	Every plan
Kimi K2.6Moonshot AI	Moonshot AI	136	Same as reading	636	Every plan
Kimi K3Moonshot AI	Moonshot AI	545	Same as reading	2,727	Every plan
Kimi K3.1 ProMoonshot AI	Moonshot AI	364	Same as reading	1,818	Every plan
GLM-5.2Z.ai	Z.ai	136	Same as reading	436	Every plan
GLM-5.3Z.ai	Z.ai	280	47.27	880	Every plan
GLM-5.3 FlashZ.ai	Z.ai	30	6	99.99	Every plan
MiniMax M3MiniMax	MiniMax	50.9	Same as reading	200	Every plan
MiniMax M3.1 FlashMiniMax	MiniMax	57.27	13.36	229	Every plan
SonarPerplexity	Perplexity	182	Same as reading	182	Every plan
Sonar ProPerplexity	Perplexity	545	Same as reading	2,727	Every plan
Sonar Reasoning ProPerplexity	Perplexity	364	Same as reading	1,454	Every plan
Amazon Nova 2 LiteAmazon	Amazon	54.54	Same as reading	455	Every plan
Amazon Nova ProAmazon	Amazon	145	Same as reading	582	Every plan
MiMo V2.6 FlashXiaomi	Xiaomi	25.45	Same as reading	50.9	Every plan
MiMo V2.6 ProXiaomi	Xiaomi	78.17	Same as reading	158	Every plan
Seed 2.0 CodeByteDance	ByteDance	90.9	Same as reading	545	Every plan
Seed 2.0 ProByteDance	ByteDance	90.9	Same as reading	545	Every plan
Hunyuan 4 (preview)Tencent	Tencent	152	Same as reading	455	Every plan
Autobench BalancedAutobench	Autobench	21.44	2.14	107	Trial, Plus and Ultra
Autobench CostAutobench	Autobench	42.87	Same as reading	129	Trial, Plus and Ultra
Autobench QualityAutobench	Autobench	429	42.87	2,144	Trial, Plus and Ultra

A million tokens is about 750,000 words. Reading is what a model takes in, From cache is context it has already seen, and Writing is what it answers. Some models cost more once a single prompt gets very long. Included in says which plans run a model: Hobby runs the open models after the trial.

Green: a small open model with few parameters at work for each token, so it uses less energy.