Live benchmark, data as of 2 September 2026

AI model speed, measured in real agent sessions

Every published speed number is one prompt, timed after the first token. This page reports what Claude and Codex models actually deliver inside working sessions: output tokens per wall-clock minute, from 118,896 measured requests across our own Claude Code and Codex fleet, by model, by week and by hour.

118,896
requests measured
7 July 2026 to 2 September 2026
6
models charted
500 or more requests each
4,311
fastest median, tokens/min
Claude Fable 5.1, 72 a second
3,628
most-used model, tokens/min
Claude Opus 5, 64,675 requests

Right now

Snapshot of 2 September 2026: medians over 118,896 requests. Live figures load when sessions are running.

Claude Opus 5
3,628
tokens/min, 60/s
Claude Fable 5
3,375
tokens/min, 56/s
Codex (GPT-5.6)
2,490
tokens/min, 42/s
Codex auto-review
1,538
tokens/min, 26/s

Observed output speed by model, main sessions

Output tokens per wall-clock minute per request. Bar spans the 10th to 90th percentile, the dot is the median. 7 July 2026 to 2 September 2026.

1,00017/s2,00033/s3,00050/s4,00067/s5,00083/s6,000100/sClaude Fable 5.14,311Claude Opus 53,628Claude Fable 53,375Claude Opus 4.83,261Codex (GPT-5.6)2,490Codex auto-review1,538

Models with fewer than 500 measured requests in a scope are in the download, not the chart. Source: Jordan James Media session transcripts, provider-billed tokens.

Claude Fable 5.1

#claude-fable-5-1
4,311tokens per minute median, 72 a second

The middle 80 percent of 1,003 requests ran between 3,050 and 5,282 tokens a minute; the aggregate rate (total tokens over total minutes) was 4,697. First content block after a median 10.862 seconds. 39 percent of its output was thinking.

Claude Opus 5

#claude-opus-5
3,628tokens per minute median, 60 a second

The middle 80 percent of 64,675 requests ran between 1,952 and 4,754 tokens a minute; the aggregate rate (total tokens over total minutes) was 3,717. First content block after a median 8.085 seconds. 18 percent of its output was thinking.

Claude Fable 5

#claude-fable-5
3,375tokens per minute median, 56 a second

The middle 80 percent of 20,282 requests ran between 1,540 and 4,760 tokens a minute; the aggregate rate (total tokens over total minutes) was 3,556. First content block after a median 8.757 seconds. 21 percent of its output was thinking.

Claude Opus 4.8

#claude-opus-4-8
3,261tokens per minute median, 54 a second

The middle 80 percent of 4,024 requests ran between 1,588 and 4,298 tokens a minute; the aggregate rate (total tokens over total minutes) was 3,532. First content block after a median 11.347 seconds. 4 percent of its output was thinking.

Codex (GPT-5.6)

#codex-gpt-5-6
2,490tokens per minute median, 42 a second

The middle 80 percent of 18,368 requests ran between 943 and 3,639 tokens a minute; the aggregate rate (total tokens over total minutes) was 2,581. First content block after a median 3.516 seconds. 37 percent of its output was thinking.

Codex auto-review

#codex-auto-review
1,538tokens per minute median, 26 a second

The middle 80 percent of 4,465 requests ran between 865 and 2,185 tokens a minute; the aggregate rate (total tokens over total minutes) was 1,493. First content block after a median 2.531 seconds. 41 percent of its output was thinking.

Subagent runs

The same models spawned as short-lived workers with a small context. Fewer than 300 requests are left out.

ModelRequestsMedian tokens/minTokens/sFirst output
Claude Sonnet 52,2544,114695.83 s
Claude Opus 52,3683,724624.566 s
Claude Haiku 4.53393,621609.26 s
Claude Fable 58353,017504.958 s

Weekly median speed, main sessions

One panel per model so the lines never cross. Shared scale, 2,000 to 4,200 tokens per minute. Weeks with fewer than 50 requests are omitted.

Claude Opus 5+11% over 7 weeks
3,0003,865W30W36
Claude Fable 5+53% over 7 weeks
3,0003,863W28W36
Codex (GPT-5.6)+6% over 3 weeks
3,0002,287W34W36
Claude Opus 4.8-29% over 6 weeks
3,0002,734W28W35

Median speed by hour of day (UTC), main sessions

Darker is faster within each model's own range of well-sampled hours. Outlined cells hold fewer than 500 requests and are left out of the shading range. Hover any hour for the value.

Claude Opus 53,415 to 3,713 tokens/min
00 UTC (10:00 Sydney)0612 (22:00 Sydney)1823
Codex (GPT-5.6)2,129 to 2,751 tokens/min
00 UTC (10:00 Sydney)0612 (22:00 Sydney)1823

Hours are UTC. Sydney is UTC plus 10; New York is UTC minus 4 in September.

How one request is timed

A real Claude Fable 5.1 reply from the session that built this page: 1,988 output tokens in 26.7 seconds, 4,467 tokens a minute. The tool then ran for 63 seconds, which is not counted.

waiting and thinking: 25.7 srequest sentanchor record written, 0 sfirst block at 25.7 sthinking, 1,850 tokenslast block at 26.7 stext and tool call streamed in 1.0 s, timer stopsthe tool then ran for 63 s, outside the window

Methodology 1.0

What the number is, how it is produced, and what it is not.

The metric

Output tokens divided by wall-clock minutes for one model request. The request starts at the transcript record written immediately before it (the user message, tool result or attachment it went out with) and ends when its last streamed content block was written. Thinking and time to first token are inside the window. Tool execution is outside it, because tool results are written after the tool finishes.

Sources

Claude Code session transcripts under ~/.claude/projects/ and Codex rollouts under ~/.codex/sessions/ on our own workstation, every record timestamped, every reply carrying the provider's usage counters. 3,595 files at the last snapshot. Main interactive sessions and subagent runs are reported separately. One measuring point in Sydney, so every first-output figure includes that network path.

Exclusions, counted

Replies under 20 output tokens (a short tool call over a few seconds says nothing about generation), placeholder and fallback turns, replies with no counters, and requests longer than an hour. Every exclusion is counted by reason in the JSON download, so a zero is a measured zero.

Units and what it is not

Tokens are provider-billed tokens, the ones invoices and rate limits count; on the same Claude text they run about 1.6 times the OpenAI o200k count that benchmark sites use. This is observed throughput, never a rate-limit quota, so it is written out in full and never abbreviated. Medians and 90th percentiles are reported, never means. Models with fewer than 500 requests in a scope stay in the download only.

Change log. 1.0 (2 September 2026): first public release, eight weeks of data, Claude Code and Codex. The live feed refreshes every minute from running sessions; the history is re-snapshotted monthly from the transcripts.

Data and API

Rows carry model names, ISO weeks, counts and durations. No project, prompt, session identifier or dollar figure is in the data.

Cite as: Jordan James Media, "AI model speed in real agent sessions", dataset of 2 September 2026, methodology 1.0, https://jordanjamesmedia.com/ai-model-speed

Free to download and reuse with a link back to this page.

Questions people ask

How fast is Claude Opus 5 in a real coding session?

Across 64,675 of our main-session requests between 7 July and 2 September 2026, Claude Opus 5 delivered a median 3,628 output tokens per wall-clock minute (60 tokens a second) with a 90th percentile of 4,754, and its first content block arrived a median 8.1 seconds after the request was sent. Those are provider-billed tokens measured end to end, including thinking, not a benchmark decode rate.

Is tokens per minute the same as the TPM rate limit?

No. Anthropic, OpenAI, Azure and Google all quote rate-limit quotas in tokens per minute, which is how many tokens you are allowed to send or receive. The figures on this page are observed throughput, how many output tokens a model actually delivered per minute inside real sessions. We write the metric out in full and never abbreviate it so the two are not confused.

Why is my agent session slower than the benchmark number?

Benchmarks time one prompt after the first token arrives and count tokens in OpenAI's o200k tokenizer. A session request also waits for the first token and for thinking, and the billed tokens you see run about 1.6 times the o200k count for the same Claude output. Put Opus 5's 3,628 billed tokens a minute in benchmark units and it is about 37 tokens a second end to end against a published 52 decode-only.

Is fast mode worth it for AI agents?

Fast mode raises decode speed by up to 2.5 times on both Anthropic and OpenAI, at 2 to 2.5 times the price or credit burn, and it does not shorten time to first token or tool execution. A request that waits eight seconds and streams for twenty goes from 28 seconds to about 16, a 1.75 times gain, and the tool runs between requests do not change. It pays for long generation with few tool calls and not for sessions dominated by waiting.

How do I measure model speed in my own Claude Code or Codex sessions?

Every Claude Code transcript under ~/.claude/projects/ and every Codex rollout under ~/.codex/sessions/ carries a timestamp on each record and the provider's usage counters on each reply. Anchor each request to the record written just before it, end it at its last streamed block, divide output tokens by minutes, drop replies under 20 tokens, and report the median and 90th percentile per model. Our measurement module and dataset are linked on this page.

Which AI coding model is fastest right now?

In our main sessions Claude Fable 5.1 has the highest median at 4,311 tokens a minute, but 39 percent of that is thinking, so Claude Opus 5 writes more visible tokens per minute. Claude Sonnet 5 is the fastest subagent at 4,114. Codex on GPT-5.6 has the lowest rate at 2,490 but the shortest wait for a first token at 3.5 seconds. Speed drifted by 11 to 20 percent within weeks, so check the live figure on this page before deciding.

The launch article
AI Model Speed, Measured: Tokens Per Minute in Real Sessions

Why benchmarks time a sprint while your agent runs a marathon, how the same model runs at three speeds, and how to choose a model on speed that is real.

Read it
Build agents you can measure
We instrument every agent fleet we build from the first session

Custom platforms, integrations and agent workflows, built and run by an AI-native studio in Sydney. This page is what that instrumentation looks like from the outside.

Talk to us about AI development