Anthropic's Claude API and OpenAI's API charge identical list prices at three tiers as of October 16, 2026: $10 input and $50 output per million tokens for Claude Fable 5.1 and GPT-6 Astra, $2 and $10 for Claude Sonnet 5.5 and GPT-6.1 Sol, and $0.10 and $0.50 for Claude Haiku 5.5 and GPT-6 Luna.
That is the published rate card on Anthropic's pricing page and OpenAI's API pricing page, both re-checked on October 16, 2026. When the sticker prices match, the bill is decided by four things the headline rate hides: how many tokens each tokenizer counts for the same text, how deep the cache discount runs, what happens above a long-prompt threshold, and whether you can wait for batch. This page lays out all four, adds Google's Gemini rates for context, and runs one workload through every model.
Claude vs OpenAI pricing: the October 2026 rate card
Claude vs OpenAI pricing is a tie on list price at the top, middle and bottom tiers in October 2026. Claude Fable 5.1 and GPT-6 Astra both cost $10/$50 per million input/output tokens, Claude Sonnet 5.5 and GPT-6.1 Sol both cost $2/$10, and Claude Haiku 5.5 and GPT-6 Luna both cost $0.10/$0.50. Only Claude Opus 5.5, at $4/$20, sits in a slot OpenAI does not fill.
| Model | Provider | Input / 1M | Cached input / 1M | Output / 1M | Note |
|---|---|---|---|---|---|
| Claude Fable 5.1 | Anthropic | $10.00 | $0.25 | $50.00 | Top general-availability Claude tier |
| GPT-6 Astra | OpenAI | $10.00 | $1.00 | $50.00 | OpenAI flagship |
| Claude Opus 5.5 | Anthropic | $4.00 | $0.20 | $20.00 | No OpenAI equivalent price point |
| Gemini 4 Argon | $2.00 → $4.00 | $0.10 → $0.20 | $10.00 → $20.00 | Promo through Nov 30, 2026 | |
| Claude Sonnet 5.5 | Anthropic | $2.00 | $0.10 | $10.00 | Same list price as GPT-6.1 Sol |
| GPT-6.1 Sol | OpenAI | $2.00 | $0.10 | $10.00 | Same list price as Sonnet 5.5 |
| Gemini 3.8 Flash | $0.75 → $1.50 | $0.075 → $0.15 | $3.75 → $7.50 | Promo through Dec 31, 2026 | |
| Gemini 3.5 Flash-Lite | $0.30 | n/a | $2.50 | Google's low-cost text tier | |
| Claude Haiku 5.5 | Anthropic | $0.10 | $0.01 | $0.50 | Prompts up to 100K tokens |
| GPT-6 Luna | OpenAI | $0.10 | $0.01 | $0.50 | OpenAI's cheapest current model |
How we counted: standard (non-batch, short-context) list prices per 1 million tokens in USD; current text models only; 10 models from 3 providers; as of October 16, 2026. Arrows show Google's scheduled price after its promotion ends. Sources: Anthropic pricing, OpenAI API pricing, Gemini API pricing.
Output price per 1M tokens, standard tier (October 2026)
Anthropic and OpenAI match at $50, $10 and $0.50; Claude Opus 5.5 at $20 is the only off-grid price
Anthropic, OpenAI and Google API pricing pages, retrieved October 16, 2026
Data
| Item | Value |
|---|---|
| Claude Fable 5.1 | $50.00 |
| GPT-6 Astra | $50.00 |
| Claude Opus 5.5 | $20.00 |
| Claude Sonnet 5.5 | $10.00 |
| GPT-6.1 Sol | $10.00 |
| Gemini 4 Argon (promo) | $10.00 |
| Gemini 3.8 Flash (promo) | $3.75 |
| Gemini 3.5 Flash-Lite | $2.50 |
| Claude Haiku 5.5 | $0.50 |
| GPT-6 Luna | $0.50 |
Why does the same list price produce different bills?
Anthropic states on its pricing page that Claude 4.7 and later models use a newer tokenizer that "produces approximately 30% more tokens for the same text," with the exact increase depending on the content. Every current Claude model in the table above falls under that note. One read on this: a prompt that OpenAI bills as 1,000 tokens could be billed as roughly 1,300 on Claude Sonnet 5.5, so the matched $2 rate behaves more like $2.60 per million OpenAI-equivalent tokens. That is Value Add VC's arithmetic on Anthropic's stated average, not a measured figure for any specific workload, and OpenAI does not publish a comparable tokenizer ratio.
Cache pricing pushes the other way. Anthropic bills cache hits on Claude Opus 5.5 and Claude Sonnet 5.5 at 0.05x the base input price ($0.20 and $0.10 per million), and on Claude Fable 5.1 at 0.025x ($0.25), against the standard 0.1x on older models, per its prompt caching table. OpenAI charges $1.00 per million cached input tokens on GPT-6 Astra, so a heavily cached agent loop on the top tier pays four times more for cache reads on OpenAI than on Claude Fable 5.1. At the middle tier both charge $0.10.
Anthropic also charges for cache writes: 1.25x base input for a five-minute cache and 2x for a one-hour cache ($2.50 and $4.00 per million on Sonnet 5.5). OpenAI's table lists cache writes too, at $12.50 per million on GPT-6 Astra and $2.50 on GPT-6.1 Sol.
How much does a real workload cost on each API?
To make the rate card concrete, take one million requests that each send 2,000 input tokens and get 500 output tokens back (2 billion input tokens and 500 million output tokens in total, no caching, standard tier). The table applies each provider's list price. The last column adds Anthropic's approximate 30% tokenizer overhead to the Claude rows; it is an estimate, labeled as such.
| Model | Input cost | Output cost | List-price total | With ~30% Claude tokenizer overhead |
|---|---|---|---|---|
| Claude Fable 5.1 | $20,000 | $25,000 | $45,000 | ~$58,500 |
| GPT-6 Astra | $20,000 | $25,000 | $45,000 | n/a |
| Claude Opus 5.5 | $8,000 | $10,000 | $18,000 | ~$23,400 |
| Gemini 4 Argon (after Dec 1) | $8,000 | $10,000 | $18,000 | n/a |
| Claude Sonnet 5.5 | $4,000 | $5,000 | $9,000 | ~$11,700 |
| GPT-6.1 Sol | $4,000 | $5,000 | $9,000 | n/a |
| Gemini 3.8 Flash (promo) | $1,500 | $1,875 | $3,375 | n/a |
| Gemini 3.5 Flash-Lite | $600 | $1,250 | $1,850 | n/a |
| Claude Haiku 5.5 | $200 | $250 | $450 | ~$585 |
| GPT-6 Luna | $200 | $250 | $450 | n/a |
How we counted: 1,000,000 requests x (2,000 input + 500 output tokens) at standard list prices as of October 16, 2026; 10 models; Value Add VC arithmetic on the providers' published rates. The tokenizer column multiplies Claude totals by 1.3 using Anthropic's stated approximate figure; real overhead varies by content.
Same workload, list price vs tokenizer-adjusted (USD)
Value Add VC arithmetic on Anthropic and OpenAI pricing pages, October 16, 2026
Data
With no caching, the tokenizer note is enough to make Claude roughly 30% dearer for the same text at every matched tier; deep cache hits can reverse that on the top tier.
What changes for long prompts, batch jobs and data residency?
| Pricing rule | Anthropic (Claude) | OpenAI | Google (Gemini) |
|---|---|---|---|
| Batch discount | 50% off input and output (Sonnet 5.5: $1/$5) | Batch tier at half price (GPT-6.1 Sol: $1/$5) | 50% cost reduction on Batch API |
| Long prompts | Full 1M-token context at standard price on Claude 4.6+ (except Haiku 5.5) | Long-context column doubles input: Astra $20, Sol $4, Luna $0.20 | Gemini 3.1 Pro: $4 input above 200K tokens vs $2 |
| Small-model long prompts | Haiku 5.5 rises to $0.50/$2.50 above 100K tokens | Luna long context $0.20/$0.75 | Flash-Lite flat $0.30/$2.50 |
| Faster output tier | Opus 5.5 fast mode: $8/$40 | Fast tier: Astra $20/$100; Ultrafast $60/$300 | Priority tier listed separately |
| Data residency | 1.1x for US-only inference on Claude 4.6+ | 10% uplift on regional endpoints for models released since Mar 5, 2026 | Not compared here |
| Cache read multiplier | 0.05x on Opus/Sonnet 5.5; 0.025x on Fable 5.1 | Astra 0.1x ($1.00); Sol 0.05x ($0.10) | Argon $0.10 (implicit caching only) |
Source: Anthropic pricing (batch, long context, fast mode, data residency sections), OpenAI pricing (Batch, Fast and Ultrafast tabs; regional-processing note), Gemini pricing; all retrieved October 16, 2026.
The long-prompt rows are where the matched rate card splits hardest. Anthropic bills a 900,000-token prompt on Claude Sonnet 5.5 at the same $2 per million as a 9,000-token one. OpenAI's long-context column charges $4 per million input on GPT-6.1 Sol, so document-heavy retrieval jobs can cost up to twice as much on input there. Claude Haiku 5.5 is the exception on Anthropic's side: its price is set per request by prompt length, and a request over 100,000 tokens pays five times the base rate.
Where does Gemini fit in the price comparison?
As of October 16, 2026, Google is the only one of the three selling its flagship below its long-run price. Google's pricing page lists Gemini 4 Argon at $2 input and $10 output per million tokens through November 30, 2026, then $4 and $20 from December 1, which would put it level with Claude Opus 5.5 rather than Sonnet 5.5. Gemini 3.8 Flash costs $0.75/$3.75 through December 31, 2026, then $1.50/$7.50 from January 1, 2027.
Two more Gemini rules matter for cost models: Gemini 4 Argon supports implicit caching only (no explicit cache control), and Google Search grounding is free for 5,000 requests a month across Gemini 3 and newer models, then $14 per 1,000 requests. For a broader view of where these labs sit on capability, see the 2026 AI model rankings.
How has Claude and OpenAI pricing changed since mid-2026?
When this page first ran in June 2026, the comparison was lopsided: Claude Opus 4.x listed at $15/$75 on the older models and the GPT-5 generation at $1.25/$10. Anthropic's price list still shows Claude Opus 4.1 at $15/$75 (retired except on Bedrock and Google Cloud), against $4/$20 for Claude Opus 5.5 now, a 73% cut in input price at the Opus tier. Claude Sonnet 5 launched with $2/$10 introductory pricing through August 31, 2026; Anthropic's pricing page now says that rate is the standard price and the scheduled rise to $3/$15 will not happen.
On OpenAI's side, GPT-6 Astra now heads the flagship table at $10/$50, while GPT-5.6 Sol carries promotional pricing at $4/$20 that OpenAI says runs at least through November 21, 2026. The full OpenAI history, including deprecation dates for GPT-4o and o3, is in our OpenAI API pricing breakdown, and the choice among Claude tiers is covered in Claude Opus vs Sonnet vs Haiku.
Which API should a startup pick on cost?
Claude is cheaper when
- A large fixed system prompt or codebase is cached and re-read many times (0.05x reads on Sonnet and Opus 5.5)
- Prompts run past OpenAI's short-context band on mid or top tiers (no long-context surcharge on Claude 4.6+)
- You need a tier between $2 and $10 input: Opus 5.5 at $4/$20 has no OpenAI twin
OpenAI is cheaper when
- Traffic is short, uncached chat or classification, where Claude's ~30% tokenizer overhead shows up directly
- Small-model prompts exceed 100K tokens (Haiku 5.5 rises to $0.50 input; Luna's long-context input is $0.20)
- You can pay for speed by request (Fast and Ultrafast tiers on Astra and Sol)
For a seed-stage team spending under a few thousand dollars a month, the tie on list price means model quality on your own evals should decide, not the rate card. Above that, the rules in this article (cache depth, prompt length, tokenizer overhead) move the bill by a third or more. Gemini 4 Argon is the cheapest flagship-class option until November 30, 2026, but a cost model built on its promotional rate needs a December line item at double the price.
What the headline misses
List prices are the least stable numbers in this market. Three of the rates on this page are scheduled to change within 11 weeks (Gemini 4 Argon, Gemini 3.8 Flash and GPT-5.6 Sol's promotion), and Anthropic reversed a planned Sonnet price rise in 2026. Per-token parity also says nothing about tokens per task: a model that solves a coding task in one pass with 3,000 output tokens beats a cheaper-looking one that takes three tries. Enterprise contracts, committed-use discounts and cloud-marketplace pricing on Amazon Bedrock or Google Cloud can differ from these first-party rates. Treat this table as the starting point for your own measurement, not the answer.
Prices verified against each provider's pricing page on October 16, 2026. For the companies behind these models, see Anthropic, OpenAI and the AI spending tracker; for what AI companies are worth relative to revenue, see AI startup valuation multiples. Published by Value Add VC · Editor-in-chief: Trace Cohen.
Latest from the Pulse
Get VC data most people never see
— free to subscribe
Trace's notes on venture, AI and startups, a few times a week. Join 5,000+ subscribers. No spam.
No spam, unsubscribe anytime. Sent via Substack · Privacy · or subscribe on Substack ↗