AI API prices don't follow the leaderboard: #1 Claude Opus 5.5 costs $4 / $20, while #3 and #4 cost $10 / $50

The model at the top of ArtificialWatch's bench index is not the most expensive one. Claude Opus 5.5 leads at 98.1 and lists at $4 per million input tokens and $20 per million output tokens. The next two models with a public price, OpenAI's GPT-6 Astra (third, 92.6) and Anthropic's own Fable 5.1 (fourth, 89.1), list at $10 / $50, two and a half times as much. Gemini 4 Carbon, in second place, has no public price yet.

$2 / $10 IS THE GOING RATE

Six of the 33 priced models on the index list at exactly $2 / $10, four of them in the top 11:

- GPT-6.1 Sol (OpenAI), 5th - Claude Sonnet 5.5 (Anthropic), 6th - Gemini 4 Argon (Google), 7th - Kimi K3.1 Pro (Moonshot), 11th - GPT-6 Sol (OpenAI), 13th - Claude Sonnet 5 (Anthropic), 37th

Gemini 4 Argon's price doubles to $4 / $20 on December 1, Google says.

THE CHEAPEST NEAR THE TOP

GPT-6.1 Terra, at $1 / $5, is the cheapest model in the top 10 (ninth, 75). Claude Haiku 5.5, at $0.10 / $0.50, is the cheapest in the top 20 (17th, 54.9). OpenAI's GPT-6 Luna lists at the same $0.10 / $0.50 and ranks 30th.

OLDER MODELS COST MORE THAN THEIR SUCCESSORS

Claude Opus 5 (10th) and Opus 4.8 (32nd) list at $5 / $25, more than Opus 5.5. OpenAI's GPT-5.5 lists at $5 / $30 (23rd) and ChatGPT-5.6 at $4 / $20 (19th), against $2 / $10 for GPT-6.1 Sol in fifth.

LONG PROMPTS COST EXTRA AT SOME LABS

OpenAI charges more for prompts over 272,000 tokens: GPT-6 Astra goes to $20 / $75, GPT-6.1 Sol and GPT-6 Sol to $4 / $15, and GPT-6.1 Terra to $2 / $7.50. xAI's Grok 4.7 goes from $2 / $6 to $4 / $12 above 200,000 tokens. Claude Haiku 5.5 rises to $0.50 / $2.50 above 100,000 tokens. ArtificialWatch's price table lists no long-prompt rate for Claude Opus 5.5 or Sonnet 5.5 across their 1M-token context.

THE FULL RANGE

Across all 40 models in ArtificialWatch's price table, input prices run from $0.10 to $10 per million tokens and output prices from $0.28 to $50.

HOW THESE ARE COUNTED

Prices are each lab's own list price on its own API, per million tokens, as read on November 3: the standard tier, without batch, cache or fast-mode rates. Ranks are from ArtificialWatch's bench index, the median of a model's percentile ranks on up to seven public leaderboards, refreshed the same day. The full table, with context windows and long-prompt rates, is at artificialwatch.com/llm-pricing, and as open data at artificialwatch.com/data/llm-prices.csv.