Home / Research / AI Token Pricing Comparison

Market research · September 2026

AI Token Pricing Comparison: September 2026

Published rate cards miss token burn. Across the four models in this study, burn varies by more than 3x. Two models at the same $50.00 per million output differ by 22 percent on a long document and by 63 percent across a coding session.

The study covers GPT-6 Astra (OpenAI), Claude Fable 5.1 (Anthropic), Gemini 3.1 Pro (Google; API id gemini-3.1-pro-preview, shortened in this report), and Grok 4.6 (xAI), each at default reasoning through the vendor's first-party API. The 240-task suite is split evenly across agentic coding, long-document analysis, multi-step research, and structured extraction. Each call logged input, cached input, visible output, and billed reasoning, and was priced on published rates as of September 1, 2026. Rates are standard synchronous unless a table says otherwise.

What the study covers

Published price is the rate. Token burn is the quantity billed to finish a task. Session cost, workload cost, and monthly spend apply both figures to larger units of work.

Published token prices

The table lists each model's published rate per million tokens, its context window, and its long-context rate. Google and xAI set the long-context threshold at 200,000 tokens and bill the higher rate against every token in the request once a prompt reaches that line. OpenAI publishes long-context rates without a published threshold; this study applied the same 200,000-token line.

Insights

• GPT-6 Astra and Claude Fable 5.1 match on input at $10.00, output at $50.00, and cache writes at $12.50 per million, and differ on cache-hit reads.
• Cached input is the widest published gap: Claude Fable 5.1 at 2.5 percent of base input, Grok 4.6 at 25 percent.
• Anthropic is the only one of the four with a flat rate through 1M, which decides cost on requests above 200,000 tokens.

Token consumption per task

The rate card sets the price of a token. The task sets how many tokens are billed. The table reports average output per task across the 240-task suite, split into visible output and billed reasoning, with a burn multiple indexed to GPT-6 Astra at 1.00x.

Insights

• Reasoning is 68.0 to 80.7 percent of billed output.
• Claude Fable 5.1 spends 3.14x the output tokens of GPT-6 Astra at the same $50.00 per million, so comparable finished work lands at an effective $157.14 per million.
• Burn lifts Grok 4.6 from $6.00 to an effective $9.36 per million and Gemini 3.1 Pro from $12.00 to $13.86.

Cumulative session cost across a 50-turn agentic run

Agentic runs resend accumulated working context as input on each turn. The table tracks cumulative spend across a 50-turn coding session of 1,030,500 input tokens, with a 72 percent prefix cache hit from the third turn. OpenAI and Anthropic bill cache writes above base input, and those writes are priced in. Google and xAI publish no separate cache-write line.

The chart plots the same figures. The four session totals separate on the first turn and remain apart through turn 50. By turn 10 the spread is $1.44. By turn 50 it is $8.45, on a session that produced the same work in every case. Input grows from about 5,400 tokens on turn 1 to about 35,800 on turn 50. Caching reduces the input side of the bill and leaves per-turn output burn as the larger remaining variable. Gemini 3.1 Pro and Grok 4.6 stay within four cents for the run. A caching-disabled control on the same session raised session totals 71 to 106 percent.

Effective cost per 1,000 runs by workload

Rate and burn combine differently by the shape of the job. The table prices four workloads at 1,000 runs each, with the task held constant across models. The customer support reply and structured classification rows ran with reasoning disabled, so output length matches and only the rate card varies. Structured classification assumes batch wherever the vendor offers it.

Insights

• Ranking changes by workload. On jobs under 200,000 tokens, GPT-6 Astra is the cheaper of the two premium models. Above that line, OpenAI's long-context rates (2x input, 1.5x output) make it the more expensive, and Anthropic stays flat.
• Full repository review puts Claude Fable 5.1 37.2 percent below GPT-6 Astra on the same 340,000-token job.
• Grok 4.6 finishes 53.4 percent above Gemini 3.1 Pro on the batched classification run because grok-4.6 does not support xAI's Batch API.

Monthly spend per developer

The figures below are monthly API spend per developer across 34 engineering teams, each on a single primary model, with the share of that spend from re-sent context and from system prompts and tools.

Insights

• The spread between the lowest and highest median is 7.5x, narrower than the 8.3x spread in published output rates.
• Re-sent context is 58 to 69 percent of every bill.
• The 90th percentile runs 3.0 to 3.8 times the median.

Conclusion

A published rate is one input to the bill. On this suite, billed reasoning, prefix caching, and the 200,000-token long-context line move session cost, the cost of 1,000 runs, and monthly spend per developer across a range the rate card alone does not show.

Sources

1. Frontier Model Token Economics Study. AIR Research Team. September 2026. Houston, TX.
2. OpenAI, API Pricing. September 2026. San Francisco, California. https://developers.openai.com/api/docs/pricing
3. Anthropic, Pricing. September 2026. San Francisco, California. https://platform.claude.com/docs/en/about-claude/pricing
4. Google, Gemini Developer API Pricing. September 2026. Mountain View, California. https://ai.google.dev/gemini-api/docs/pricing
5. xAI, Models and Pricing. September 2026. Palo Alto, California. https://docs.x.ai/docs/pricing
6. Artificial Analysis, Independent Analysis of AI Models. September 2026. Sydney, Australia. https://artificialanalysis.ai/models

← All research