LLM Token Pricing Comparison — Reasoning-Capable Channels

DOWNLOAD PDF reference sheet —

Take this card into the boardroom.

The same table, formatted for printing and sharing. PDF reference sheet, one page, ready for procurement conversations.

File: llm-pricing-sep-2026-em-reference.pdf · ~106 KB · 1 page

The Numbers

All rates USD per 1M tokens. Cache rates noted where published.

Channel	Model	Input /1M	Output /1M	Notes
OpenAI	GPT-6 Astra flagship	$10.00	$50.00	Frontier
OpenAI	GPT-6 Sol	$2.00	$10.00	Cached in $0.20
OpenAI	GPT-6 Luna	$0.10	$0.50	Cached in $0.01
Anthropic	Claude Fable 5 frontier	$10.00	$50.00
Anthropic	Claude Opus 5	$5.00	$25.00
Anthropic	Claude Sonnet 5	$2.00	$10.00
Anthropic	Claude Haiku 4.5	$1.00	$5.00
Google	Gemini 3.1 Pro	$2.00	$12.00	≤200K ctx; $4/$18 >200K
Google	Gemini 3.7 Flash	$0.75	$3.75	Intro thru Dec 31, then $1.50/$7.50
Google	Gemini 2.5 Flash-Lite	$0.10	$0.40	Thinking tokens bill as output
DeepSeek	V4-Flash	$0.14	$0.28	Cache hit $0.0028; off-peak ≈ half
DeepSeek	V4-Pro	$0.435	$0.87	Cache hit $0.0036
Z.ai	GLM-5.3-Flash	$0.15	$0.50	Cached in $0.03; 1M ctx; natively multimodal
Z.ai	GLM-5.3	$1.40	$4.40	Cached in $0.26
MiniMax	MiniMax-M3	$0.30	$1.20	≤512K ctx; $0.60/$2.40 >512K; cache $0.06
Moonshot AI	Kimi K3	$3.00	$15.00	1M ctx; cache-hit in $0.30; no reasoning surcharge
Moonshot AI	Kimi K2.7 Code	$0.95	$4.00	256K ctx; batch −40%
Moonshot AI	Kimi K2.5	$0.60	$3.00	Value tier
OpenRouter	(router — pass-through)	provider rate	+ 5.5% credit fee	BYOK 1M reqs/mo free; some models free
Ollama Cloud	Pro plan	$20/mo flat	—	Usage-credit pools + GPU-time quotas

All rates USD per 1M tokens (input / output). Cache rates listed where published by the vendor. Verify against each provider's current pricing page before contracting.

Reading the Numbers

Cheapest real reasoning per token: DeepSeek V4-Flash ($0.14/$0.28) and GPT-6 Luna ($0.10/$0.50) — frontier-family reasoning at flash pricing.
Batch APIs: −50% on OpenAI, Anthropic, and Google; −40% on Moonshot (excluding K3). Caching: −90% on input (Anthropic), ~99.9% on cache hits (DeepSeek), $0.03 (GLM Flash), $0.30 (Kimi K3).
Watch: Gemini bills thinking tokens at output rates; DeepSeek official docs now show a peak/off-peak split (peak = 2×).
OpenRouter is a router, not a model: per-token price = underlying provider rate + fee. Value = failover + one API key.
Verify before contracting. Prices as published late September 2026. Provider pricing pages move; the PDF is a snapshot, not a contract.

Sources

OpenAI pricing page and 23 September launch posts; Anthropic, Google, DeepSeek, Z.ai, MiniMax, and Moonshot official pricing via aggregators, September 2026.

Compiled by Executive Mind · executivemind.io