Two published price rises land before January, and both are Google's. If Gemini 4 Argon is in your stack, your
December invoice is twice your November one unless you move. Current rates for every major model are below,
followed by every change with a date attached.
THE SHORT VERSION
● Cheapest input: Claude Haiku 5.5 at $0.10 per million
● The $2/$10 cluster: Gemini 4 Argon, Claude Sonnet 5.5 and GPT-6.1 Sol all sit here
● Most expensive: GPT-6 Astra at $10 input and $50 output
● 1 December 2026: Gemini 4 Argon doubles to $4/$20 - this is the one to budget for
● 1 January 2027: Gemini 3.8 Flash rises to $1.50/$7.50
Current rates, per million tokens
| Model | Provider | Input | Output | Note |
|---|
| Claude Haiku 5.5 | Anthropic | $0.10 | not stated | Cheapest listed input rate |
| Gemini 4 Argon | Google | $2.00 | $10.00 | Introductory - rises 1 Dec 2026 |
| Claude Sonnet 5.5 | Anthropic | $2.00 | $10.00 | Standard list |
| GPT-6.1 Sol | OpenAI | $2.00 | $10.00 | Cached input $0.10 |
| Claude Opus 5.5 | Anthropic | $4.00 | $20.00 | Standard list |
| Meta Muse Wonder | Meta | $4.00 | $20.00 | Standard list |
| GPT-6 Astra | OpenAI | $10.00 | $50.00 | Most expensive listed |
Standard API list prices. Batch, cached and priority tiers differ and are not shown. Read 26 October 2026.
Scheduled changes
| Date | Model | From | To | What is happening |
|---|
| 1 Dec 2026 | Gemini 4 Argon | $2 / $10 | $4 / $20 | Introductory period ends; Batch, Flex and Priority tiers double too |
| 1 Jan 2027 | Gemini 3.8 Flash | promotional | $1.50 / $7.50 | Promotional rate ends 31 Dec 2026 |
What the table actually tells you
The interesting structure is the cluster. Gemini 4 Argon, Claude Sonnet 5.5 and GPT-6.1 Sol all sit at
exactly $2 input and $10 output. Three vendors, three different models, identical pricing - that is a
market that has converged on a price point for the capable-but-not-flagship tier, not a coincidence. It also means
switching between them is a capability decision rather than a cost one.
Above that, Claude Opus 5.5 and Meta Muse Wonder both sit at $4/$20, exactly double the cluster.
GPT-6 Astra is the outlier at $10/$50 - five times the cluster rate and two and a half times the
$4/$20 tier. Astra tops several capability boards, so the question it poses is whether the margin over Opus 5.5
justifies 2.5x the cost on your workload specifically.
Gemini 4 Argon's position in that cluster is temporary. It is there on an introductory rate that expires, and from
1 December it joins the $4/$20 tier alongside Opus 5.5 and Muse Wonder. Google also doubles the Batch, Flex and
Priority tiers on the same date, so there is no cheaper lane to retreat into.
Cached input is where the real difference sits
Headline rates get quoted; cached rates decide the bill on anything with a long system prompt. Gemini 4 Argon
caches input at $0.10 per million, a 95% discount on its introductory rate. GPT-6.1 Sol caches at
$0.10 as well, half what GPT-6 Sol charged. If your workload re-sends the same context
repeatedly, that number matters more than the input rate above it.
Frequently asked questions
What is the cheapest AI API per million tokens?
Of the models listed here, Claude Haiku 5.5 at $0.10 per million input tokens. Among capable mid-tier models, Gemini 4 Argon, Claude Sonnet 5.5 and GPT-6.1 Sol all sit at $2 input and $10 output.
When does Gemini 4 Argon get more expensive?
1 December 2026. It rises from $2/$10 to $4/$20 per million tokens when the introductory period ends on 30 November, and the Batch, Flex and Priority tiers double at the same time.
How much does GPT-6.1 Sol cost?
$2.00 per million input tokens, $10.00 per million output tokens, and $0.10 per million for cached input - half GPT-6 Sol's cached rate.
Why is GPT-6 Astra so much more expensive?
At $10 input and $50 output it is five times the $2/$10 cluster. It ranks at or near the top of several capability boards, so the question is whether that margin justifies the cost on your specific workload.
Is Gemini 3.8 Flash changing price?
Yes. Its promotional rate ends 31 December 2026 and it rises to $1.50 per million input and $7.50 per million output on 1 January 2027.
Do these prices include batch or cached rates?
No. The table shows standard API list prices only. Batch, cached and priority tiers are priced differently by every vendor, and cached input in particular can be a 95% discount.
Sources
Gemini 4 Argon pricing, introductory period and comparative list prices from VentureBeat, 13 October 2026. GPT-6.1 Sol rates from OpenAI. Gemini 3.8 Flash change from the 9to5AI schedule. See also our deprecation tracker.
Last updated 26 October 2026. Dates and prices change - verify against the vendor before acting.