Lab concentration of routed tokens (OpenRouter, HHI)

Herfindahl index of prompt-plus-completion tokens by lab across the models in OpenRouter's daily top table, each model counted toward the lab in its OpenRouter id, last reading each week. Derived metric lab_token_hhi. One router's traffic and the listed slice only.

Reason through this

The reading

The series starts in the week of 7 September 2026; a direction needs five weekly readings. On 9 September the top table spread across twelve labs, led by Chinese open-weight authors.

Direction over the last 5 readingseach dot links to its record

index

in the rule's windowearliermoves within ±0.020 index read as stable, and a move counts only if more than half the steps go its way; this window moved +0.056 index and 2 of 4 steps rose; higher reads concentrating

measured

9 readings drawn from the derived metric lab_token_hhi; 178 observations in the evidence.

Derived (lab_token_hhi) from: OpenRouter rankings (token totals, top models by day). OpenRouter.

The numbers

as of	reading	value	record
2026-11-03	—	0.24 index	measured
2026-11-01	—	0.25 index	measured
2026-10-25	—	0.23 index	measured
2026-10-18	—	0.17 index	measured
2026-10-11	—	0.19 index	measured
2026-10-04	—	0.18 index	measured
2026-09-27	—	0.15 index	measured
2026-09-20	—	0.17 index	measured
2026-09-13	—	0.16 index	measured

Why it matters

Multi-homing shrinks lock-in, which shrinks the labs' bargaining share; a falling index is the leading sign of model commoditisation in buyer behaviour.

What cuts against it

OpenRouter is one router with a developer-heavy mix and only its top table is counted; enterprise buyers mostly go direct to the labs and never appear here. Ramp's letters measure a different thing (the share of US businesses paying each lab) and are not counted as a second source for this index.

How it is measured

derived metric lab_token_hhi (formula in the semantic layer)
source OpenRouter rankings (token totals, top models by day) · default tier 3 · OpenRouter terms; public page
proxy market_structure · cadence weekly

Four weekly readings; under 0.02 in the index is table churn. Rising = routed tokens concentrating on fewer labs.

Applied to metric:lab_token_hhi.

Derived rows (9)

as of	value	inputs
2026-11-03	0.24 index	sources1 2 3+17 more
2026-11-01	0.25 index	sources1 2 3+17 more
2026-10-25	0.23 index	sources1 2 3+16 more
2026-10-18	0.17 index	sources1 2 3+17 more
2026-10-11	0.19 index	sources1 2 3+16 more
2026-10-04	0.18 index	sources1 2 3+17 more
2026-09-27	0.15 index	sources1 2 3+17 more
2026-09-20	0.17 index	sources1 2 3+17 more
2026-09-13	0.16 index	sources1 2 3+17 more

Latest readings

2026-11-03 · 0.24 indexas of 2026-11-03(20 obs)
2026-11-01 · 0.25 indexas of 2026-11-01(20 obs)
2026-10-25 · 0.23 indexas of 2026-10-25(19 obs)
2026-10-18 · 0.17 indexas of 2026-10-18(20 obs)
2026-10-11 · 0.19 indexas of 2026-10-11(19 obs)
2026-10-04 · 0.18 indexas of 2026-10-04(20 obs)

Why this status, and its history

It has 3 readings and its direction rule compares 5, so it reads emerging rather than not yet measurable: the site uses not yet measurable only for what cannot be measured at all. The status turns into a direction once enough readings exist.

2026-09-27not yet measurable to emergingconf 40 → 40 · Claude (model)
It has 3 readings and its direction rule compares 5, so it reads emerging rather than not yet measurable: the site uses not yet measurable only for what cannot be measured at all. The status turns into a direction once enough readings exist.
Weighed against it: None needed: no reading has changed; this corrects the word used for too few readings.
sources1 2 3 4 5 6+14 more
2026-09-11unmeasured to not yet measurableconf — → 40 · Claude (model)
First weekly reading, 9 Sep 2026: routed tokens in OpenRouter's top table spread across twelve labs, an index of 0.16, led by Tencent, Z.ai and DeepSeek ahead of OpenAI. The direction rule needs five weekly readings; not yet measurable until then. Replaces the Ramp instrument, which had one monthly reading.
sources1 2 3 4 5 6+13 more

Why this confidence

Rests on 178 observations from OpenRouter rankings (token totals, top models by day); best evidence tier 3, grade B; current.

Confidence is independent of status: 90–95 multiple strong independent sources; 70–89 good evidence, some ambiguity; 50–69 mixed or hard to operationalise; 0–49 limited or vague evidence.

Predictions that lean on this

Both sides expect this
Use of models stays spread across several labs and open models, with the concentration of routed tokens among labs below 0.25 index, and few firms train or run their own. · Soren Larson and Garicano and Saa-Requejo

Reading

emergingA model's judgement, not a reading: probably stable. The weekly readings on the site wobble up and down and the latest sits only just above the first, a move that falls a shade inside the noise band, so they were extended as no clear movement. The level itself describes traffic spread across many labs. Against this, the last week rose sharply, and one more reading like it would tip the rule to concentrating.

Latest

0.24 indexas of 2026-11-03(20 obs)

What the rule reads

0.24 indexas of 2026-11-03(20 obs)

The rule

Direction over 5 readings; moves under 0.020 index count as stable; higher means concentrating.

Confidence 40 of 95: limited or vague evidence. grade B

Leading: developers move routed traffic between labs within days of a release, ahead of enterprise contracts and revenue.

Updated 2026-09-11