Current Claude Code figures on npm downloads, developer surveys, revenue run rates, benchmarks, and pricing, with sample sizes and measurement periods.
Links marked with * are affiliate links. If a purchase is made through such links, we receive a commission.
I’ve spent more than 800 hours working with Claude Code.
I started in mid-2025. What I can’t tell you from my own experience is how many other people use it. And Anthropic isn’t saying.
A useful clue comes from JetBrains’ 2026 survey. In it, 39% of more than 15,000 professional developers said they use Claude Code at work.
And Claude Code has only existed since February 24, 2025, when it launched as a research preview.
In this article, I’ll show you the key Claude Code statistics on usage, revenue, pricing, and benchmarks. Here’s the short version:
TL;DRKey Takeaways
In JetBrains' 2026 survey, 39% of more than 15,000 professional developers said they use Claude Code at work.
Anthropic reported an annualized Claude Code revenue run rate above $2.5 billion on February 12, 2026.
The Claude Code npm package reached 562.7 million downloads through October 6, 2026.
Note
This article covers Claude Code as a product. For figures about Claude models and Anthropic overall, see Claude Statistics 2026. Figures on OpenAI's competing product are in Codex Statistics.
1. What Is Claude Code?
Claude Code is Anthropic’s coding agent. It can read a project, edit files, run commands, and manage changes with Git. According to the official documentation, Anthropic offers it in the terminal, IDEs, a desktop app, and a browser.
The main public milestones show the difference between product preview and general availability:
Feb 24, 2025
Research preview
Anthropic first introduced Claude Code as a limited preview.
May 2025
Generally available
Anthropic made Claude Code generally available in May.
The GitHub repository had 154,017 stars according to its repository page on October 31, 2026.
154,017 GitHub stars on October 31, 2026
562,748,153 npm downloads from February 24, 2025, to October 6, 2026
52,946,380 npm downloads in the 30-day period from September 7 to October 6, 2026
Version 2.1.316, released October 30, 2026, according to the GitHub releases
The 30-day figure comes from the npm download API. According to npm, these are HTTP downloads of package archives, including build servers, mirrors, and bots.
Anthropic has not published an absolute count of active Claude Code users. The 2026 Developer Ecosystem Survey from JetBrains offers a current comparison of reported use at work. Its tenth edition surveyed more than 15,000 professional developers worldwide. Across all AI coding agents, 90% of respondents reported using one at least weekly and 68% daily.
The Pragmatic Engineer survey ran from January 27 to February 17, 2026, among newsletter readers, with 906 respondents. A total of 507 of them, or 56%, regularly used AI agents.
4.1 Tools Used by Regular Agent Users
Among regular agent users, 71% named Claude Code, well ahead of GitHub Copilot and Cursor.
Surveys are snapshots. The gradually.ai popularity index tracks demand month by month instead. It combines Google search volume, usage data such as downloads and installs, and news mentions into a score from 0 to 100.
Popularity index from 0 to 100, higher is more popular · As of November 9, 2026
Anthropic publishes some revenue figures as run rates. The chart shows the Claude Code values. August and February are lower bounds (“more than”), and November is the announced $1 billion milestone. Bloomberg describes the metric as projecting a shorter period across a full year.
The figures for Anthropic overall are in a separate table. They do not cover Claude Code alone:
February 12, 2026
Scope
Anthropic overall
Reported figure
$14 billion run rate
Source type
Anthropic statement
Early May 2026
Scope
Anthropic overall
Reported figure
More than $47 billion run rate
Source type
Anthropic statement on May 28
End of July 2026
Scope
Anthropic overall
Reported figure
More than $65 billion run rate
Source type
Bloomberg report, anonymous sources
September 2026
Scope
Anthropic overall
Reported figure
More than $100 billion annualized revenue expected for 2026
Source type
NYT report, cited by Axios
Scope
Reported figure
Source type
February 12, 2026
Anthropic overall
$14 billion run rate
Anthropic statement
Early May 2026
Anthropic overall
More than $47 billion run rate
Anthropic statement on May 28
End of July 2026
Anthropic overall
More than $65 billion run rate
Bloomberg report, anonymous sources
September 2026
Anthropic overall
More than $100 billion annualized revenue expected for 2026
NYT report, cited by Axios
For Claude Code alone, the latest separately stated run rate I found in a public company source was more than $2.5 billion on February 12, 2026. In Anthropic’s February 12 announcement, the company said it had more than doubled since the start of the year. It also said weekly active Claude Code users had doubled since January 1.
Opus 5.5 was released on September 22, 2026, and Sonnet 5.5 on September 28. For Opus 5.5, Anthropic’s Opus release report specifies Xhigh effort and a 2.6 percentage point standard error.
According to the Sonnet system card, the Sonnet test used Claude Code in “--bare” mode at maximum effort, across 66 tasks and five trials per task, without internet access. The standard error is 2.5 percentage points.
7. What We Know About Successful Work and Productivity
An Anthropic study published June 16, 2026 analyzed about 400,000 interactive sessions from approximately 235,000 people between October 2025 and April 2026. It covered Claude Code in the terminal, Claude.ai, and the desktop app. The table shows shares of sessions with code changes.
A session counted as a verified success only if a classifier judged its goal to have been met and there was a verifiable signal such as a commit or passing tests. Sessions without a clear goal (about 7.7%) were excluded.
The same study assigns every session to one of nine work modes. A little over half involve writing, fixing, and testing code. Another 17% concern operating software, and a bit more than a quarter planning, understanding, analysis, and documents.
In Anthropic's October 16, 2026 study on work inside its own company, code commits per engineer rose to five times their level six months earlier. The study also follows 192 engineers who adopted coding agents against 74 who did not. Adoption was not random, and more code does not automatically mean better quality or business outcomes.
Anthropic’s Claude Code Analytics API gives organizations daily metrics, including accepted or rejected proposals for file-editing tools, sessions, lines changed, commits, and pull requests.
8. Pricing and Costs
Claude Code is not included in Claude’s free plan. Individual access starts with Pro. Max and Ultra offer higher usage limits.
8.1 Monthly Plan Prices
The chart shows US list prices per month with monthly billing, before taxes. Team prices are per seat.
Anthropic states plan usage only relative to Pro. Max (formerly Max 5x) offers five times the Pro usage per 5-hour session and per week. Ultra (formerly Max 20x) has no 5-hour limit for verified accounts and offers ten times the Pro weekly usage. Anthropic does not publish absolute message or hour counts. All paid plans also have a weekly limit.
According to Anthropic’s pricing page, Pro costs $200 up front with annual billing, which works out to about $17 per month. Team and Enterprise also include Claude Code.
Team requires at least two seats. With annual billing, a Standard seat costs $20 and a Premium seat $100 per month. Premium seats offer five times the usage of Standard seats. Enterprise costs $20 per seat per month, billed annually with a minimum of 20 seats. Usage is billed on top at API rates.
8.3 API Prices per Million Tokens
API pricing depends on the model and number of input and output tokens. Anthropic lists these standard rates per million tokens for its generally available Claude models.
Input
Output
USD per 1M tokens
Claude Fable 5.1
Input: $10
Output: $50
Claude Opus 5.5
Input: $4
Output: $20
Claude Sonnet 5.5
Input: $2
Output: $10
Claude Haiku 5.5 (two price tiers, bar shows the first)
Input: $0.1 / $0.5
Output: $0.5 / $2.5
Claude Fable 5
Input: $10
Output: $50
Claude Opus 5
Input: $5
Output: $25
Claude Sonnet 5
Input: $2
Output: $10
Claude Opus 4.7
Input: $5
Output: $25
Claude Opus 4.6
Input: $5
Output: $25
Claude Sonnet 4.6
Input: $3
Output: $15
Claude Haiku 4.5
Input: $1
Output: $5
Source: Anthropic API pricing documentation
|
CC BY 4.0
gradually.ai
According to Anthropic’s API pricing documentation, prompt caching and Opus Fast mode have separate prices.
In the gradually.ai popularity index, Claude Code scores 93 out of 100. That is rank 1 of the 36 AI harnesses we track. The index combines search demand, GitHub and package data and mentions in tech media.
In the EU, Claude Code is at rank 1 of 31, in Germany at rank 1 of 31. Use the switch above the chart to change between world, EU and Germany.
The most popular AI harnesses
Popularity index from 0 to 100, higher is more popular.
As of November 9, 2026 · Source: gradually.ai popularity index
11. How popular are Anthropic's models?
Anthropic's most popular model is Claude Opus 5.5 with 86 out of 100 (rank 4 of 645 AI models with a popularity score). In total, 16 Anthropic models have a score; the chart highlights them.
In the EU, Claude Opus 5.5 is at rank 10 of 613, in Germany, Claude Opus 5.5 is at rank 8 of 692. Use the switch above the chart to change between world, EU and Germany.
Popularity of Anthropic models
Popularity index from 0 to 100, higher is more popular.
25 of 111 models
Gemini 3.8 Flash88
DeepSeek-V4.1-Flash87
GLM-5.387
Claude Opus 5.586
Kimi K386
Nano Banana 285
GLM-5.3-Flash85
DeepSeek-V4-Flash85
GPT-6 Luna85
GPT-6 Astra84
MiniMax H384
GPT-5.6 Luna83
Nano Banana Pro82
GLM-5.282
Claude Opus 581
GPT Image 281
Qwen 3.8 27B81
Claude Haiku 5.581
Claude Sonnet 581
Claude Sonnet 5.579
Claude Fable 5.156
Claude Sonnet 4.654
Claude Sonnet 4.550
Claude Fable 545
Claude Opus 4.837
020406080100
Gemini 3.8 Flash, Google: 8888
DeepSeek-V4.1-Flash, DeepSeek: 8787
GLM-5.3, Z.ai: 8787
Claude Opus 5.5, Anthropic: 8686
Kimi K3, Moonshot AI: 8686
Nano Banana 2, Google: 8585
GLM-5.3-Flash, Z.ai: 8585
DeepSeek-V4-Flash, DeepSeek: 8585
GPT-6 Luna, OpenAI: 8585
GPT-6 Astra, OpenAI: 8484
MiniMax H3, MiniMax: 8484
GPT-5.6 Luna, OpenAI: 8383
Nano Banana Pro, Google: 8282
GLM-5.2, Z.ai: 8282
Claude Opus 5, Anthropic: 8181
GPT Image 2, OpenAI: 8181
Qwen 3.8 27B, Alibaba: 8181
Claude Haiku 5.5, Anthropic: 8181
Claude Sonnet 5, Anthropic: 8181
Claude Sonnet 5.5, Anthropic: 7979
Claude Fable 5.1, Anthropic: 5656
Claude Sonnet 4.6, Anthropic: 5454
Claude Sonnet 4.5, Anthropic: 5050
Claude Fable 5, Anthropic: 4545
Claude Opus 4.8, Anthropic: 3737
Gemini 3.8 Flash
DeepSeek-V4.1-Flash
GLM-5.3
Claude Opus 5.5
Kimi K3
Nano Banana 2
GLM-5.3-Flash
DeepSeek-V4-Flash
GPT-6 Luna
GPT-6 Astra
MiniMax H3
GPT-5.6 Luna
Nano Banana Pro
GLM-5.2
Claude Opus 5
GPT Image 2
Qwen 3.8 27B
Claude Haiku 5.5
Claude Sonnet 5
Claude Sonnet 5.5
Claude Fable 5.1
Claude Sonnet 4.6
Claude Sonnet 4.5
Claude Fable 5
Claude Opus 4.8
Source: gradually.ai popularity index
12. How good are Anthropic's models?
In the gradually.ai overall index, which combines independent benchmarks into one value, Claude Fable 5.2 is Anthropic's strongest model: rank 1 of 94 with 130.9 points. 15 Anthropic models are ranked in the index. 100 points is the average of a fixed reference group; 15 points more means a model wins about three of four benchmark comparisons.
Overall index (Gradually Meta-Score v5)
24 of 94 models
4 models in this view are measured at a lower reasoning effort than max in the overall index. Dashed column: lower reasoning effort than the top.
Artificial Analysis uses AA-Omniscience to measure how often a model gives a wrong answer to a knowledge question instead of admitting it does not know. Anthropic's best result comes from Claude Haiku 4.5: 23% of its answers that are not correct are wrong guesses (rank 12 of 66, rank 1 hallucinates least). The median of all measured models is 39%.
7 Anthropic models have a measurement. Each model counts with its best measured setting. The rate describes reliability, not safety in the sense of jailbreaks or harmful content.
Hallucination rate
Share of wrong answers among all answers that are not correct, lower is better. Best measured setting per model.
26 of 66 models
Gemini 4 Carbon4%
Gemini 4 Argon4%
Command A Plus14%
Muse Ember 1.017%
Muse Wonder Full18%
Qwen 3.8 27B18%
Grok 4.720%
Muse Wonder21%
Solar Pro 421%
Muse Spark 1.322%
Nemotron 3 Ultra23%
Claude Haiku 4.523%
Mistral Large 425%
DeepSeek V4.1 Flash 101725%
GLM-5.226%
Granite 4.2 30B27%
Gemini 3.5 Flash-Lite27%
Qwen 3.8 Max 090227%
Muse Spark Open28%
K2 Horizon 375B-A23B29%
Claude Haiku 5.534%
Claude Sonnet 5.543%
Claude Opus 5.550%
Claude Fable 5.256%
Claude Opus 559%
Claude Fable 5.160%
020406080100
Gemini 4 Carbon, Google DeepMind: 4%4%
Gemini 4 Argon, Google DeepMind: 4%4%
Command A Plus, Cohere: 14%14%
Muse Ember 1.0, Meta: 17%17%
Muse Wonder Full, Meta: 18%18%
Qwen 3.8 27B, Alibaba: 18%18%
Grok 4.7, xAI: 20%20%
Muse Wonder, Meta: 21%21%
Solar Pro 4, Upstage: 21%21%
Muse Spark 1.3, Meta: 22%22%
Nemotron 3 Ultra, NVIDIA: 23%23%
Claude Haiku 4.5, Anthropic: 23%23%
Mistral Large 4, Mistral AI: 25%25%
DeepSeek V4.1 Flash 1017, DeepSeek: 25%25%
GLM-5.2, Z.ai: 26%26%
Granite 4.2 30B, IBM: 27%27%
Gemini 3.5 Flash-Lite, Google: 27%27%
Qwen 3.8 Max 0902, Alibaba: 27%27%
Muse Spark Open, Meta: 28%28%
K2 Horizon 375B-A23B, Institute of Foundation Models: 29%29%
Claude Haiku 5.5 is the fastest with 164 output tokens per second. The median of all measured models is 69 tokens per second. OpenRouter has measured 14 Anthropic models.
Speed
28 of 94 models
Nemotron 3.5 Lightning180tokens/s · Median across providers on OpenRouter from the readings of the last 24 hours (each covering the last 30 minutes), excluding time to first token · Retrieved 2026-11-08
Claude Haiku 5.5164tokens/s · Median across providers on OpenRouter from the readings of the last 24 hours (each covering the last 30 minutes), excluding time to first token · Retrieved 2026-11-08
Gemini 3.5 Flash-Lite145tokens/s · Median across providers on OpenRouter from the readings of the last 24 hours (each covering the last 30 minutes), excluding time to first token · Retrieved 2026-11-08
Gemini 3.5 Flash142tokens/s · Median across providers on OpenRouter from the readings of the last 24 hours (each covering the last 30 minutes), excluding time to first token · Retrieved 2026-11-08
GPT-5 nano136tokens/s · Median across providers on OpenRouter from the readings of the last 24 hours (each covering the last 30 minutes), excluding time to first token · Retrieved 2026-11-08
GPT-5.4 mini134tokens/s · Median across providers on OpenRouter from the readings of the last 24 hours (each covering the last 30 minutes), excluding time to first token · Retrieved 2026-11-08
DeepSeek V4.1 Flash 1017126tokens/s · Median across providers on OpenRouter from the readings of the last 24 hours (each covering the last 30 minutes), excluding time to first token · Retrieved 2026-11-08
Gemini 3.1 Flash-Lite125tokens/s · Median across providers on OpenRouter from the readings of the last 24 hours (each covering the last 30 minutes), excluding time to first token · Retrieved 2026-11-08
GPT-5.6 Luna123tokens/s · Median across providers on OpenRouter from the readings of the last 24 hours (each covering the last 30 minutes), excluding time to first token · Retrieved 2026-11-08
Mistral Medium 3.5118tokens/s · Median across providers on OpenRouter from the readings of the last 24 hours (each covering the last 30 minutes), excluding time to first token · Retrieved 2026-11-08
DeepSeek-V4.1-Flash118tokens/s · Median across providers on OpenRouter from the readings of the last 24 hours (each covering the last 30 minutes), excluding time to first token · Retrieved 2026-11-08
Mistral Small 4112tokens/s · Median across providers on OpenRouter from the readings of the last 24 hours (each covering the last 30 minutes), excluding time to first token · Retrieved 2026-11-08
Gemini 3.6 Flash110tokens/s · Median across providers on OpenRouter from the readings of the last 24 hours (each covering the last 30 minutes), excluding time to first token · Retrieved 2026-11-08
GPT-4.1 nano110tokens/s · Median across providers on OpenRouter from the readings of the last 24 hours (each covering the last 30 minutes), excluding time to first token · Retrieved 2026-11-08
Ling 3.0 Flash106tokens/s · Median across providers on OpenRouter from the readings of the last 24 hours (each covering the last 30 minutes), excluding time to first token · Retrieved 2026-11-08
GPT-5 mini105tokens/s · Median across providers on OpenRouter from the readings of the last 24 hours (each covering the last 30 minutes), excluding time to first token · Retrieved 2026-11-08
Claude Sonnet 5.5103tokens/s · Median across providers on OpenRouter from the readings of the last 24 hours (each covering the last 30 minutes), excluding time to first token · Retrieved 2026-11-08
Muse Spark 1.3102tokens/s · Median across providers on OpenRouter from the readings of the last 24 hours (each covering the last 30 minutes), excluding time to first token · Retrieved 2026-11-08
Gemini 3 Flash Preview101tokens/s · Median across providers on OpenRouter from the readings of the last 24 hours (each covering the last 30 minutes), excluding time to first token · Retrieved 2026-11-08
GLM-5.3-Flash101tokens/s · Median across providers on OpenRouter from the readings of the last 24 hours (each covering the last 30 minutes), excluding time to first token · Retrieved 2026-11-08
Claude Haiku 4.584tokens/s · Median across providers on OpenRouter from the readings of the last 24 hours (each covering the last 30 minutes), excluding time to first token · Retrieved 2026-11-08
Claude Sonnet 577tokens/s · Median across providers on OpenRouter from the readings of the last 24 hours (each covering the last 30 minutes), excluding time to first token · Retrieved 2026-11-08
Claude Opus 5.575tokens/s · Median across providers on OpenRouter from the readings of the last 24 hours (each covering the last 30 minutes), excluding time to first token · Retrieved 2026-11-08
Claude Opus 4.871tokens/s · Median across providers on OpenRouter from the readings of the last 24 hours (each covering the last 30 minutes), excluding time to first token · Retrieved 2026-11-08
Claude Opus 571tokens/s · Median across providers on OpenRouter from the readings of the last 24 hours (each covering the last 30 minutes), excluding time to first token · Retrieved 2026-11-08
Claude Fable 5.159tokens/s · Median across providers on OpenRouter from the readings of the last 24 hours (each covering the last 30 minutes), excluding time to first token · Retrieved 2026-11-08
Claude Opus 4.757tokens/s · Median across providers on OpenRouter from the readings of the last 24 hours (each covering the last 30 minutes), excluding time to first token · Retrieved 2026-11-08
Claude Fable 550tokens/s · Median across providers on OpenRouter from the readings of the last 24 hours (each covering the last 30 minutes), excluding time to first token · Retrieved 2026-11-08
tokens/s050100150200
Nemotron 3.5 Lightning, NVIDIA: 180 tokens per second · Median across providers on OpenRouter from the readings of the last 24 hours (each covering the last 30 minutes), excluding time to first token · Retrieved 2026-11-08180
Claude Haiku 5.5, Anthropic: 164 tokens per second · Median across providers on OpenRouter from the readings of the last 24 hours (each covering the last 30 minutes), excluding time to first token · Retrieved 2026-11-08164
Gemini 3.5 Flash-Lite, Google: 145 tokens per second · Median across providers on OpenRouter from the readings of the last 24 hours (each covering the last 30 minutes), excluding time to first token · Retrieved 2026-11-08145
Gemini 3.5 Flash, Google: 142 tokens per second · Median across providers on OpenRouter from the readings of the last 24 hours (each covering the last 30 minutes), excluding time to first token · Retrieved 2026-11-08142
GPT-5 nano, OpenAI: 136 tokens per second · Median across providers on OpenRouter from the readings of the last 24 hours (each covering the last 30 minutes), excluding time to first token · Retrieved 2026-11-08136
GPT-5.4 mini, OpenAI: 134 tokens per second · Median across providers on OpenRouter from the readings of the last 24 hours (each covering the last 30 minutes), excluding time to first token · Retrieved 2026-11-08134
DeepSeek V4.1 Flash 1017, DeepSeek: 126 tokens per second · Median across providers on OpenRouter from the readings of the last 24 hours (each covering the last 30 minutes), excluding time to first token · Retrieved 2026-11-08126
Gemini 3.1 Flash-Lite, Google: 125 tokens per second · Median across providers on OpenRouter from the readings of the last 24 hours (each covering the last 30 minutes), excluding time to first token · Retrieved 2026-11-08125
GPT-5.6 Luna, OpenAI: 123 tokens per second · Median across providers on OpenRouter from the readings of the last 24 hours (each covering the last 30 minutes), excluding time to first token · Retrieved 2026-11-08123
Mistral Medium 3.5, Mistral AI: 118 tokens per second · Median across providers on OpenRouter from the readings of the last 24 hours (each covering the last 30 minutes), excluding time to first token · Retrieved 2026-11-08118
DeepSeek-V4.1-Flash, DeepSeek: 118 tokens per second · Median across providers on OpenRouter from the readings of the last 24 hours (each covering the last 30 minutes), excluding time to first token · Retrieved 2026-11-08118
Mistral Small 4, Mistral AI: 112 tokens per second · Median across providers on OpenRouter from the readings of the last 24 hours (each covering the last 30 minutes), excluding time to first token · Retrieved 2026-11-08112
Gemini 3.6 Flash, Google: 110 tokens per second · Median across providers on OpenRouter from the readings of the last 24 hours (each covering the last 30 minutes), excluding time to first token · Retrieved 2026-11-08110
GPT-4.1 nano, OpenAI: 110 tokens per second · Median across providers on OpenRouter from the readings of the last 24 hours (each covering the last 30 minutes), excluding time to first token · Retrieved 2026-11-08110
Ling 3.0 Flash, inclusionAI: 106 tokens per second · Median across providers on OpenRouter from the readings of the last 24 hours (each covering the last 30 minutes), excluding time to first token · Retrieved 2026-11-08106
GPT-5 mini, OpenAI: 105 tokens per second · Median across providers on OpenRouter from the readings of the last 24 hours (each covering the last 30 minutes), excluding time to first token · Retrieved 2026-11-08105
Claude Sonnet 5.5, Anthropic: 103 tokens per second · Median across providers on OpenRouter from the readings of the last 24 hours (each covering the last 30 minutes), excluding time to first token · Retrieved 2026-11-08103
Muse Spark 1.3, Meta: 102 tokens per second · Median across providers on OpenRouter from the readings of the last 24 hours (each covering the last 30 minutes), excluding time to first token · Retrieved 2026-11-08102
Gemini 3 Flash Preview, Google: 101 tokens per second · Median across providers on OpenRouter from the readings of the last 24 hours (each covering the last 30 minutes), excluding time to first token · Retrieved 2026-11-08101
GLM-5.3-Flash, Z.ai: 101 tokens per second · Median across providers on OpenRouter from the readings of the last 24 hours (each covering the last 30 minutes), excluding time to first token · Retrieved 2026-11-08101
Claude Haiku 4.5, Anthropic: 84 tokens per second · Median across providers on OpenRouter from the readings of the last 24 hours (each covering the last 30 minutes), excluding time to first token · Retrieved 2026-11-0884
Claude Sonnet 5, Anthropic: 77 tokens per second · Median across providers on OpenRouter from the readings of the last 24 hours (each covering the last 30 minutes), excluding time to first token · Retrieved 2026-11-0877
Claude Opus 5.5, Anthropic: 75 tokens per second · Median across providers on OpenRouter from the readings of the last 24 hours (each covering the last 30 minutes), excluding time to first token · Retrieved 2026-11-0875
Claude Opus 4.8, Anthropic: 71 tokens per second · Median across providers on OpenRouter from the readings of the last 24 hours (each covering the last 30 minutes), excluding time to first token · Retrieved 2026-11-0871
Claude Opus 5, Anthropic: 71 tokens per second · Median across providers on OpenRouter from the readings of the last 24 hours (each covering the last 30 minutes), excluding time to first token · Retrieved 2026-11-0871
Claude Fable 5.1, Anthropic: 59 tokens per second · Median across providers on OpenRouter from the readings of the last 24 hours (each covering the last 30 minutes), excluding time to first token · Retrieved 2026-11-0859
Claude Opus 4.7, Anthropic: 57 tokens per second · Median across providers on OpenRouter from the readings of the last 24 hours (each covering the last 30 minutes), excluding time to first token · Retrieved 2026-11-0857
Claude Fable 5, Anthropic: 50 tokens per second · Median across providers on OpenRouter from the readings of the last 24 hours (each covering the last 30 minutes), excluding time to first token · Retrieved 2026-11-0850
Via the API, Anthropic's cheapest model, Claude Haiku 5.5, costs $0.20 per 1M tokens. The most expensive is Claude Fable 5 at $20.00. Both are blended prices of three parts input to one part output, as in the LLM price index.
API prices
Blended price in US dollars per 1M tokens, 3 parts input to 1 part output
At Sonnet 5.5's list rates of $2 per million input tokens and $10 per million output tokens, 10,000 all-input tokens cost $0.02 and all-output tokens $0.10. For 200,000 tokens, that is $0.40 or $2.00. A real session mixes token types, and Claude Code subscriptions include the usage in the plan.
Anthropic has not published a current absolute count of active Claude Code users. In JetBrains' weighted Developer Ecosystem Survey 2026, 39% of more than 15,000 professional developers worldwide said they used Claude Code at work in May-July 2026.
No. The 39% figure from the JetBrains survey for May-July 2026 refers to professional developers worldwide who self-reported using Claude Code at work. It does not measure revenue share or paid licenses.
More than $2.5 billion in run-rate revenue, as Anthropic reported on February 12, 2026. A run rate annualizes the current revenue pace.
No. In its October 16, 2026 study “How employment is changing at Anthropic”, Anthropic reported that code commits per engineer had risen to five times their level six months earlier. Separately, the study compared 192 engineers who adopted coding agents with 74 who did not. Anthropic cautions that adoption was not random and that more code does not necessarily mean better quality or business outcomes.
Under US pricing, Pro costs $20 with monthly billing or $200 paid up front for a year (effectively $17 per month). Max costs $100 per month and offers five times the Pro usage. Ultra costs $200 per month and offers ten times the Pro weekly usage. API usage is billed separately by model and token count.
Finn Hillebrandt is the founder of Gradually AI, an SEO and AI expert. He helps online entrepreneurs simplify and automate their processes and marketing with AI. Finn shares his knowledge here on the blog in 50+ articles as well as through the AI Business Club.
gradually.ai
