Large Language Models

LLM platforms compared

The products you actually buy. Each one wraps one or more of the models in the next table.

Showing 8 of 8 platforms

	Flagship model
DeepSeek Cheapest input	DeepSeek V4	1M	64K	≈$0.44 / 1M input · ≈$0.87 / 1M output 1M context under an MIT license. Cache-hit input bills substantially lower than the cache-miss rate quoted here.	Free
Mistral Le Chat	Mistral Large 3	256K	32K	$0.50 / 1M input · $1.50 / 1M output EU data residency and on-prem deployment are available, which is usually the deciding factor over raw price.	From $15
Gemini	Gemini 3.1 Pro	1M	64K	$2.00 / 1M input · $12.00 / 1M output (≤200K context) Above 200K context the rate steps to $4.00 input / $18.00 output. Context caching cuts repeat-prompt cost by up to ~90%.	From $20
Grok	Grok 4.6	500K	64K	$2.00 / 1M input · $6.00 / 1M output (<200K prompt) Once a prompt reaches 200K tokens the entire request bills at $4.00 input / $12.00 output. Cached input is $0.50 / 1M.	From $30
ChatGPT Widest context	GPT-5.6 Sol	1.05M	128K	$5.00 / 1M input · $30.00 / 1M output Flagship reasoning tier. Output pricing is the highest of the mainstream closed flagships at this input rate.	From $20
Claude	Claude Opus 5	1M	128K	$5.00 / 1M input · $25.00 / 1M output 1M context at standard pricing with no long-context premium. Prompt caching reads bill at roughly 0.1x.	From $20
Perplexity	GPT-5.6 Sol	1.05M	128K	$5.00 / 1M input · $30.00 / 1M output Flagship reasoning tier. Output pricing is the highest of the mainstream closed flagships at this input rate.	From $20
Microsoft Copilot	GPT-5.6 Sol	1.05M	128K	$5.00 / 1M input · $30.00 / 1M output Flagship reasoning tier. Output pricing is the highest of the mainstream closed flagships at this input rate.	From $20

API prices are per 1M tokens and reflect each platform's flagship tier. Cheaper tiers exist on most platforms — see the individual reviews. Verified 13 August 2026.

Every model, by price

The model layer underneath, sorted cheapest input price first. The spread between the cheapest and most expensive model here is over 300x per input token — capability differences are real, but they are nowhere near 300x.

Model	Creator	Context	Max output	API price / 1M tokens	Weights
Qwen3.7 Flash	Alibaba	1M	32K	$0.03 / 1M input · $0.13 / 1M output	OpenApache 2.0
DeepSeek V4 Flash	DeepSeek	1M	32K	$0.14 / 1M input · $0.28 / 1M output	OpenMIT
Mistral Small 4	Mistral AI	256K	32K	$0.15 / 1M input · $0.60 / 1M output	OpenApache 2.0
GPT-5.6 Luna	OpenAI	1.05M	128K	$0.20 / 1M input · $1.20 / 1M output	Closed
Llama 4 Maverick 400B (MoE, 17B active)	Meta	1M	32K	≈$0.24 / 1M input · ≈$0.97 / 1M output	OpenLlama 4 Community License
Qwen3.5 Plus	Alibaba	1M	32K	$0.40 / 1M input · $2.40 / 1M output	OpenApache 2.0
DeepSeek V4	DeepSeek	1M	64K	≈$0.44 / 1M input · ≈$0.87 / 1M output	OpenMIT
Mistral Large 3	Mistral AI	256K	32K	$0.50 / 1M input · $1.50 / 1M output	OpenMistral Research / Commercial License
Gemini 3.6 Flash	Google	1M	64K	$0.75 / 1M input · $3.75 / 1M output	Closed
Gemini 3.7 Flash	Google	1M	64K	$0.75 / 1M input · $3.75 / 1M output	Closed
Kimi K2.6	Moonshot AI	256K	32K	≈$0.95 / 1M input · ≈$4.00 / 1M output	OpenModified MIT
Claude Haiku 4.5	Anthropic	200K	64K	$1.00 / 1M input · $5.00 / 1M output	Closed
Gemini 2.5 Pro	Google	1M	64K	$1.25 / 1M input · $10.00 / 1M output (≤200K context)	Closed
Grok 4.3	xAI	1M	64K	$1.25 / 1M input · $2.50 / 1M output	Closed
GLM-5.2	Zhipu AI	200K	32K	≈$1.40 / 1M input · ≈$4.40 / 1M output	OpenMIT
Gemini 3.1 Pro	Google	1M	64K	$2.00 / 1M input · $12.00 / 1M output (≤200K context)	Closed
GPT-5.6 Terra	OpenAI	1.05M	128K	$2.00 / 1M input · $12.00 / 1M output	Closed
Grok 4.6	xAI	500K	64K	$2.00 / 1M input · $6.00 / 1M output (<200K prompt)	Closed
GPT-5.4	OpenAI	1.05M	128K	$2.50 / 1M input · $15.00 / 1M output	Closed
Claude Sonnet 5	Anthropic	1M	128K	$3.00 / 1M input · $15.00 / 1M output	Closed
Kimi K3 2.8T (MoE)	Moonshot AI	1.05M	64K	$3.00 / 1M input · $15.00 / 1M output	OpenModified MIT
Claude Opus 5	Anthropic	1M	128K	$5.00 / 1M input · $25.00 / 1M output	Closed
GPT-5.6 Sol	OpenAI	1.05M	128K	$5.00 / 1M input · $30.00 / 1M output	Closed
Claude Fable 5	Anthropic	1M	128K	$10.00 / 1M input · $50.00 / 1M output	Closed
Llama 5 600B (MoE)	Meta	5M	32K	Free to self-host · hosted rates vary by provider	OpenLlama 5 Community License

25 models tracked. Prices are list rates for the standard API route and exclude caching discounts, batch discounts, and long-context surcharges — those are noted per model in the platform reviews. Verified 13 August 2026.

What you would actually pay

The same models as a cost comparison. Output tokens run several times the input rate on almost every model, so the split matters as much as the headline number.

1M input tokens1M output tokens

Qwen3.7 Flash
Alibaba
$0.16
DeepSeek V4 Flash
DeepSeek
$0.42
Mistral Small 4
Mistral AI
$0.75
Llama 4 Maverick
Meta
$1.21
DeepSeek V4
DeepSeek
$1.30
GPT-5.6 Luna
OpenAI
$1.40
Mistral Large 3
Mistral AI
$2
Qwen3.5 Plus
Alibaba
$2.80
Grok 4.3
xAI
$3.75
Gemini 3.7 Flash
Google
$4.50
Gemini 3.6 Flash
Google
$4.50
Kimi K2.6
Moonshot AI
$4.95
GLM-5.2
Zhipu AI
$5.80
Claude Haiku 4.5
Anthropic
$6
Grok 4.6
xAI
$8
Gemini 2.5 Pro
Google
$11.25
GPT-5.6 Terra
OpenAI
$14
Gemini 3.1 Pro
Google
$14
GPT-5.4
OpenAI
$17.50
Claude Sonnet 5
Anthropic
$18
Kimi K3
Moonshot AI
$18
Claude Opus 5
Anthropic
$30
GPT-5.6 Sol
OpenAI
$35
Claude Fable 5
Anthropic
$60

Cost of one million input tokens plus one million output tokens, at list API rates. Bars share a linear scale, so Qwen3.7 Flash ($0.16) is a sliver next to Claude Fable 5 ($60) — a 375x spread. Your actual bill depends on your input/output mix; caching, batch discounts and long-context surcharges are excluded. Exact figures and their sources are in the table above. Verified 16 August 2026.