AI pricing, translated into real bills. Pick a content size and compare OpenAI, Anthropic, Gemini, xAI Grok, Groq, DeepSeek, Mistral, Qwen, OpenRouter and more.
New to tokens? Understand AI token usage and cost. Looking for a lowest-rate shortlist? See the cheapest LLM APIs. See how TokenScale verifies prices.
Looking for price history? It's built into the Calculator →
What The Hobbit costs · Mid tier · current price
Cheapest first · same checked observations as the full table below; carried rows stay labelled.
How to read it: pick a content size, tier and price basis below, or tap a preset above. “Total” shows what it costs to have the model read that content once and write that much back. Models that cannot fit the combined request are not ranked.
Tokens ≈ words × 1.33 — a reasonable average, not an exact count; actual tokens vary by model and tokenizer. Full methodology →
Batch mode uses exact model-specific rules only. Unsupported routes and routes without enough current public evidence are shown as not ranked; no universal discount is assumed. Cache pricing remains out of scope. Check the ↗ link on each row for the official page.
At 10,000 calls/day, this workload ranges from $2,536.46/day on Alibaba Qwen to $44,388.05/day on Sakana AI. Monthly: $76,093.80 to $1,331,641.50.
With $500.00/month, you can run about 65 calls/day on Alibaba Qwen, or 3 calls/day on Sakana AI.
Prefer the raw feed? Every nightly rate is machine-readable in
prices.json (append-only history; entries flagged
cf were carried forward, not verified that night) and dated promotional
overlays are in promos.json. How these are
verified: Methodology.
| # | |||
|---|---|---|---|
| 1 | Alibaba Qwen | $0.25 | cheapest |
| 2 | xAI Grok | $0.48 | ×1.9 |
| 3 | Google Gemini | $0.57 | ×2.3 |
| 4 | Moonshot Kimi | $0.63 | ×2.5 |
| 5 | DeepSeek | $0.67 | ×2.6 |
| 6 | Meta | $0.70 | ×2.7 |
| 7 | AWS Bedrock | $1.52 | ×6.0 |
| 8 | Anthropic | $1.52 | ×6.0 |
| 9 | OpenAI | $1.78 | ×7.0 |
| 10 | Sakana AI | $4.44 | ×18 |
| not ranked | Cerebras | Doesn’t fit | not ranked |
| not ranked | Cohere | Doesn’t fit | not ranked |
| not ranked | DeepInfra | Doesn’t fit | not ranked |
| not ranked | Fireworks AI | Doesn’t fit | not ranked |
| not ranked | Groq | Doesn’t fit | not ranked |
| not ranked | MiniMax | Doesn’t fit | not ranked |
| not ranked | Mistral AI | Doesn’t fit | not ranked |
| not ranked | OpenRouter | Doesn’t fit | not ranked |
| not ranked | Perplexity | Doesn’t fit | not ranked |
| not ranked | SambaNova | Doesn’t fit | not ranked |
| not ranked | Together AI | Doesn’t fit | not ranked |
| not ranked | Zhipu GLM | Doesn’t fit | not ranked |
Time-window rule: DeepSeek uses its peak on-demand rate in this comparison. Its official off-peak rate is 50% lower outside 01:00–04:00 and 06:00–10:00 UTC. TokenScale does not invent an average.
Every price change we’ve recorded, in plain terms. One row per change, with input and output shown separately. Percentage labels combine input plus output at equal token weight; model swaps remain labelled as swaps.
94 price changes recorded across 22 providers
| Now ($/M in · out) | |||
|---|---|---|---|
| 2026-10-01 | DeepInfra | $0.09 · $0.34 | ↔ model swap |
| 2026-09-29 | OpenRouter | $0.25 · $1.00 | ↑ combined +10% pricier |
| 2026-09-25 | Anthropic | $4.00 · $20.00 | ↔ model swap |
| 2026-09-25 | OpenAI | $2.00 · $10.00 | ↔ model swap |
| 2026-09-25 | OpenAI | $0.10 · $0.50 | ↔ model swap |
| 2026-09-10 | DeepSeek | $0.30 · $1.20 | ↔ model swap |
| 2026-08-29 | Fireworks AI | $1.32 · $3.96 | ↑ combined +1% pricier |
| 2026-08-18 | Groq | $0.15 · $0.60 | ↔ model swap |
| 2026-08-18 | Groq | $0.075 · $0.30 | ↔ model swap |
| 2026-08-16 | DeepSeek | $1.32 · $3.96 | ↑ combined +305% pricier |
| 2026-08-16 | DeepSeek | $1.32 · $3.96 | ↑ combined +305% pricier |
| 2026-08-16 | DeepSeek | $0.44 · $1.32 | ↑ combined +319% pricier |
| 2026-08-12 | OpenRouter | $0.24 · $0.90 | ↓ combined 11% cheaper |
| 2026-08-11 | AWS Bedrock | $0.24 · $0.97 | ↔ model swap |
| 2026-08-10 | Meta | $1.25 · $4.25 | ↔ model swap |
| 2026-08-10 | Meta | $1.25 · $4.25 | ↔ model swap |
| 2026-08-10 | Meta | $1.25 · $4.25 | ↔ model swap |
| 2026-08-10 | xAI Grok | $2.00 · $6.00 | ↔ model swap |
| 2026-08-10 | OpenRouter | $0.26 · $1.03 | ↑ combined +29% pricier |
| 2026-07-31 | OpenRouter | $0.20 · $0.80 | ↓ combined 13% cheaper |
| 2026-07-31 | DeepInfra | $0.02 · $0.04 | ↑ combined +20% pricier |
| 2026-07-30 | OpenAI | $2.00 · $12.00 | ↓ combined 20% cheaper |
| 2026-07-30 | OpenAI | $0.20 · $1.20 | ↓ combined 80% cheaper |
| 2026-07-28 | Moonshot Kimi | $3.00 · $15.00 | ↔ model swap |
| 2026-07-28 | Alibaba Qwen | $0.25 · $1.50 | ↑ combined +52% pricier |
| 2026-07-28 | Groq | $0.15 · $0.60 | ↔ model swap |
| 2026-07-28 | OpenAI | $1.00 · $6.00 | ↔ model swap |
| 2026-07-28 | Google Gemini | $1.50 · $7.50 | ↔ model swap |
| 2026-07-28 | Google Gemini | $0.30 · $2.50 | ↔ model swap |
| 2026-07-21 | OpenRouter | $0.23 · $0.91 | ↑ combined +14% pricier |
| 2026-07-21 | DeepInfra | $0.20 · $0.80 | ↑ combined +33% pricier |
| 2026-07-21 | DeepInfra | $0.02 · $0.03 | ↓ combined 29% cheaper |
| 2026-07-21 | Together AI | $1.04 · $1.04 | ↑ combined +18% pricier |
| 2026-07-03 | Alibaba Qwen | $0.17 · $0.99 | ↓ combined 34% cheaper |
| 2026-07-02 | Anthropic | $10.00 · $50.00 | ↔ model swap |
| 2026-07-01 | Meta | $0.15 · $0.60 | ↔ model swap |
| 2026-07-01 | MiniMax | $0.30 · $1.20 | ↔ model swap |
| 2026-07-01 | Moonshot Kimi | $1.90 · $8.00 | ↔ model swap |
| 2026-07-01 | Alibaba Qwen | $2.50 · $7.50 | ↔ model swap |
| 2026-07-01 | Alibaba Qwen | $0.40 · $1.60 | ↔ model swap |
| 2026-07-01 | Alibaba Qwen | $0.25 · $1.50 | ↔ model swap |
| 2026-07-01 | xAI Grok | $1.00 · $2.00 | ↔ model swap |
| 2026-07-01 | SambaNova | $3.00 · $4.50 | ↔ model swap |
| 2026-07-01 | SambaNova | $0.22 · $0.59 | ↔ model swap |
| 2026-07-01 | Cerebras | $0.35 · $0.75 | ↔ model swap |
| 2026-07-01 | Cerebras | $0.35 · $0.75 | ↔ model swap |
| 2026-07-01 | Fireworks AI | $1.74 · $3.48 | ↔ model swap |
| 2026-07-01 | Fireworks AI | $0.15 · $0.60 | ↔ model swap |
| 2026-07-01 | Fireworks AI | $0.07 · $0.30 | ↔ model swap |
| 2026-07-01 | DeepInfra | $0.15 · $0.60 | ↔ model swap |
| 2026-07-01 | Together AI | $1.20 · $4.50 | ↔ model swap |
| 2026-07-01 | Together AI | $0.88 · $0.88 | ↓ combined 15% cheaper |
| 2026-07-01 | Together AI | $0.05 · $0.20 | ↔ model swap |
| 2026-07-01 | Groq | $1.00 · $3.00 | ↔ model swap |
| 2026-07-01 | Mistral AI | $0.50 · $1.50 | ↔ model swap |
| 2026-07-01 | Mistral AI | $1.50 · $7.50 | ↔ model swap |
| 2026-07-01 | Mistral AI | $0.15 · $0.60 | ↔ model swap |
| 2026-06-23 | MiniMax | $0.30 · $1.20 | ↑ combined +20% pricier |
| 2026-06-23 | Zhipu GLM | $1.00 · $3.20 | ↑ combined +57% pricier |
| 2026-06-23 | xAI Grok | $1.25 · $2.50 | ↓ combined 53% cheaper |
| 2026-06-23 | Together AI | $1.04 · $1.04 | ↑ combined +18% pricier |
| 2026-06-18 | Zhipu GLM | $1.40 · $4.40 | ↑ combined +43% pricier |
| 2026-06-13 | Alibaba Qwen | $1.20 · $6.00 | ↑ combined +38% pricier |
| 2026-06-13 | OpenRouter | $0.10 · $0.32 | ↓ combined 19% cheaper |
| 2026-06-13 | Cerebras | $2.25 · $2.75 | ↑ combined +262% pricier |
| 2026-06-13 | AWS Bedrock | $2.40 · $2.40 | ↓ combined 77% cheaper |
| 2026-06-13 | Mistral AI | $0.20 · $0.60 | ↑ combined +100% pricier |
| 2026-06-12 | Anthropic | $5.00 · $25.00 | ↔ model swap |
| 2026-06-11 | DeepSeek | $0.43 · $0.87 | ↓ combined 75% cheaper |
| 2026-06-11 | DeepSeek | $0.43 · $0.87 | ↓ combined 75% cheaper |
| 2026-06-10 | Anthropic | $10.00 · $50.00 | ↑ combined +100% pricier |
| 2026-06-04 | OpenRouter | $0.20 · $0.80 | ↑ combined +138% pricier |
| 2026-05-25 | OpenRouter | $0.70 · $2.50 | ↑ combined +19% pricier |
| 2026-05-25 | AWS Bedrock | $1.00 · $5.00 | ↑ combined +25% pricier |
| 2026-05-20 | Perplexity | $2.00 · $8.00 | ↑ combined +67% pricier |
| 2026-05-20 | Mistral AI | $0.40 · $2.00 | ↑ combined +50% pricier |
| 2026-05-20 | Google Gemini | $1.50 · $9.00 | ↔ model swap |
| 2026-05-19 | DeepInfra | $0.10 · $0.32 | ↓ combined 33% cheaper |
| 2026-05-19 | DeepInfra | $0.02 · $0.05 | ↓ combined 36% cheaper |
| 2026-05-19 | AWS Bedrock | $3.00 · $15.00 | ↓ combined 44% cheaper |
| 2026-05-19 | DeepSeek | $1.74 · $3.48 | ↑ combined +281% pricier |
| 2026-05-15 | xAI Grok | $2.00 · $6.00 | ↔ model swap |
| 2026-05-15 | xAI Grok | $1.25 · $2.50 | ↔ model swap |
| 2026-05-15 | xAI Grok | $0.20 · $0.50 | ↓ combined 13% cheaper |
| 2026-05-15 | DeepSeek | $1.74 · $3.48 | ↔ model swap |
| 2026-05-15 | DeepSeek | $0.14 · $0.28 | ↓ combined 39% cheaper |
| 2026-05-15 | Anthropic | $5.00 · $25.00 | ↓ combined 67% cheaper |
| 2026-05-15 | OpenAI | $5.00 · $30.00 | ↑ combined +250% pricier |
| 2026-05-15 | OpenAI | $2.50 · $15.00 | ↑ combined +40% pricier |
| 2026-05-15 | OpenAI | $0.75 · $4.50 | ↑ combined +600% pricier |
| 2026-05-15 | Google Gemini | $2.00 · $12.00 | ↑ combined +24% pricier |
| 2026-05-14 | Anthropic | $3.00 · $15.00 | ↓ combined 44% cheaper |
| 2026-05-12 | OpenAI | $2.00 · $8.00 | ↓ combined 80% cheaper |
| 2026-05-12 | OpenAI | $2.50 · $10.00 | ↓ combined 50% cheaper |
The model named is the tier’s current occupant — a big jump usually means the slot switched to a different model, not that one model repriced overnight. The full story behind any move is in the Changelog. Charts of all of this: Price Charts.
The Journal, condensed. Every post as a headline — tap one for the gist, follow the link for the full story.
The nightly numbers had started to look almost suspiciously tidy. Night after night, TokenScale was accounting for all 66 tracked prices with the kind of consistency I had spent months trying to build. That should have felt… Read the full post →
After building TokenScale with AI, including a lot of it while travelling in July, I stopped deliberately adding new features and began watching the systems around it. I wanted to find out what could run without constant… Read the full post →
Today we made a small but important change to how TokenScale checks its work. Before a price update can go live, it now has to pass through two separate copies of the… Read the full post →
This morning I did what I ask every visitor to do: I looked at the grey. We ran a full verification pass today, the first complete attempt since the end of July. 42 of the 66 tracked tiers read clean at the provider's own page.… Read the full post →
Post 10 said the tool was finished, and it is. What is never finished is the record, because the record is the product now, and this week it turned three months old. That is a milestone worth marking properly: old enough that… Read the full post →
This is the last build post of TokenScale's first chapter. The tool is finished. Not abandoned, finished. And the way the final piece went in says everything about how the whole thing was made. I didn't fit it at a desk. I fitted… Read the full post →
As of today, TokenScale's daily statistics collection no longer runs on the Mac in my office. What it gathers is anonymous counting only: how many people visited, from where, and what they read, never anything about who they are.… Read the full post →
Automating a daily job isn't one build, it's a few weeks of the thing quietly teaching you what you got wrong. TokenScale's nightly price check has been settling in, and it handed me three failures in a row. Each was caught late.… Read the full post →
The changelog, condensed. Every entry as a headline, newest first — tap one for the gist.
TypeSafe AI’s Jev is a specialist decision model. It evaluates supplied state against bounded questions and returns typed choices, scores or probabilities. It does not write prose or code, so putting its price beside a model that… Read the full entry →
Google replaced its Antigravity managed-agent preview with antigravity-preview-09-2026 , deprecated the 05-2026 preview and scheduled the older preview to shut down on 5 October. xAI separately released Grok Voice Transcribe 2.0… Read the full entry →
Meta's current official pricing catalog lists Muse Spark 1.3, 1.2 and 1.1 on the same Standard basis: $1.25 input, $4.25 output and $0.15 cached input per million tokens. TokenScale still tracks Muse Spark 1.2 in all three… Read the full entry →
Every price on TokenScale is an input cost plus an output cost. Until today the calculator’s single-reply view had no way to say how long the answer was, so it assumed the answer was exactly as long as what you sent . For a text… Read the full entry →
The 18 August entry correctly recorded DeepSeek V4 Flash at $0.44 input and $1.32 output per million tokens at peak, but TokenScale omitted an important qualifier from its explanation: the 01:00–04:00 and 06:00–10:00 UTC peak… Read the full entry →
OpenAI lists GPT-5.6 Cyber as an approval-gated model for authorized cybersecurity research and testing, priced at $12.50 input and $75 output per million tokens. It is a specialist, not a general-purpose replacement for GPT-5.6… Read the full entry →
DeepSeek now publishes two standard on-demand windows rather than one flat price. TokenScale uses the highest published rate as the comparison figure and labels it as peak: V4 Flash is $0.44 input and $1.32 output per million… Read the full entry →
Alibaba still marks Qwen3.7-Max as a limited-time 50% promotion, but its public pricing page does not supply a usable start and end date. TokenScale therefore continues to record the published $2.50 input and $7.50 output list… Read the full entry →
The reliability page exists to count the nights the record missed. From 4 to 9 August it counted them wrong, and wrong in its own favour: it showed 14 carried-forward nights while the price record held 17 , and it left the August… Read the full entry →
First: the live version file, BUILD.txt , could serve a cached copy up to five days stale (it reported v514 while v518 was live). It now ships with a no-store header, so neither browsers nor the CDN may cache it. Second: the… Read the full entry →
Between 13 and 17 July the tracker stopped writing new data, and it took three nights to notice. The cause was mundane: a macOS update reset a permission the background job needs to read its own files, and it failed — quietly —… Read the full entry →
Some models get a launch. Fable 5 got a saga. Anthropic's Mythos-class flagship held our Pro slot for exactly 48 hours in June — 10th to 12th — before a US export-control directive suspended it worldwide, and the board reverted… Read the full entry →
Once in a while the right move is to stop trusting your own machinery and check everything by hand. On 1 July we ran a full drift audit: every one of the 22 providers × 3 tiers on this board, re-verified against the provider's… Read the full entry →
Anthropic launched Claude Sonnet 5 on 30 June. It steps into TokenScale's Anthropic Mid slot in place of Sonnet 4.6, at the same standard list price — $3/M in, $15/M out — and Anthropic says it closes much of the gap to Opus.… Read the full entry →
OpenAI unveiled the GPT-5.6 family on 26 June — three models named Sol , Terra and Luna , a new naming scheme that replaces the old mini/nano tiers with capability tiers inside one generation. Access is the story in itself: it's… Read the full entry →
TokenScale now records a small set of anonymous, cookieless engagement events — things like "the page was scrolled past the pricing", "the calculator was used", or "a provider was viewed" — so we can see which parts actually help… Read the full entry →
Zhipu launched GLM-5.2 on 16 June — the first MIT-licensed, 1M-context model to hold a flagship tier. TokenScale's Zhipu GLM flagship slot moves to $1.40/M in, $4.40/M out , against GPT-5.5's $5/$30 at the same context. The +43%… Read the full entry →
13 June re-priced five tiers at once, in both directions. Cerebras jumped +299% while AWS Bedrock cut hosted Llama 3.1 405B output from $16 to $2.40 — the night's biggest drop, the opposite direction on the same run. Mistral,… Read the full entry →
The 10 June note below recorded Anthropic's Pro slot moving to Fable 5 at $10/$50. That didn't hold. Between 10–12 June the Pro tier round-tripped — $5→$10→$5 in, $25→$50→$25 out — settling back at Opus 4.8 ($5/$25) . TokenScale… Read the full entry →
DeepSeek cut its Mid/Pro tier (V4 Pro) by 75% on 11 June — $1.74→$0.435 in, $3.48→$0.87 out . The relentless undercutting that's kept the Novel Index floor near half a cent. Read the full entry →
Anthropic launched Claude Fable 5 on 9 June, its first generally available Mythos-class model: a tier above Opus. TokenScale's Anthropic Pro slot now maps to Fable 5 (high) at $10/M in, $50/M out. That makes it the most expensive… Read the full entry →
DeepSeek used to be the only Chinese lab on the board. That stopped making sense. Added four open-weight frontier labs — Alibaba Qwen , Zhipu GLM , Moonshot Kimi and MiniMax — plus Meta's own first-party Llama API, so you no… Read the full entry →
The month's biggest swings now live on their own page — and each move exports as a branded card you can drop straight into a thread. A percentage tells a developer something; a whole novel going from $1.91 to $0.32 tells everyone. Read the full entry →
Google updated pricing on their mid-tier Gemini model. Silver (Flash-Lite) stayed the same. Gold (Flash 3.5) did not. Read the full entry →
Hours before posting to Hacker News, we found that our hero number was quoting input cost only. Here's what we fixed — and why it made for a better story. Read the full entry →
The "Which model should I use?" quiz was silently failing on first run. A missing DOM element meant the result screen crashed before anyone could see a recommendation. Read the full entry →
The first public version. A single HTML file, no backend, no sign-up. Pricing for 16 AI providers expressed in content you recognise. Read the full entry →
Provider pages are authoritative for current published prices. TokenScale records each nightly observation as verified or carried forward.
TokenScale · Bilton Projects · free, no sign-up · prices through 2026-10-05