Published
Updated
Inception calls Mercury Decide first on OpenRouter's JevBench v1.4, and prints no score
Opened October 8, 2026, the JevBench API board ranks Mercury Decide third at composite 72.4. The free model page returned the same day, priced Free, with up to 14 decisions a second and no JevBench score. Inception posted on September 30 that Mercury Decide is the most intelligent decision model on OpenRouter, citing JevBench v1.4, at up to 14 decisions a second. The post does not print a score. The free model URL returned Not Found. OpenRouter's Mercury 2.5 page describes Mercury Decide as a System One endpoint with free output tokens and the same schema as Jev. An October 1 search of Benchmark Heaven found no Mercury Decide.
Inception posted on September 30 that Mercury Decide is the most intelligent decision model on OpenRouter, citing JevBench v1.4, at up to 14 decisions a second. The attachment is a chess video. We did not transcribe a score from it. The post does not print a JevBench number. OpenRouter posted the same claim four minutes earlier. The free early-access URL, openrouter.ai/inception/mercury-decide:free, returned Not Found. So did openrouter.ai/inception/mercury-decide.
The Mercury 2.5 page describes Mercury Decide as a System One endpoint. A caller sends a state and typed questions. The model returns a choice, a score, or a yes/no, and the page says output tokens are free. It names the same /v1/systemone schema as Jev, and the same ceiling of 14 decisions a second. The page calls the probability calibrated. We did not measure that.
On October 1 we opened the JevBench page again. It still said v1.5.4 and 106 ranked systems. A search of that text for “Mercury Decide” returned no match. The v1.4 board on this desk is the older harmonic mean, where Jev is 63.29. The post does not print a Mercury cell on either edition. We did not rerun the board.
Opened October 8, 2026, the API board is release v1.6.1, revision v1.7.21, and ranks 26 hosted systems. Mercury Decide is third at composite 72.4. The row links to inception/mercury-decide:free. The capability card prints 73.6, Intelligence 65.9, Calibration 81.3, $0.018 per 1,000 decisions, and capability rank 5. The September 30 post still prints no score. We did not rerun the suite.
Opened the same day, openrouter.ai/inception/mercury-decide:free returned the model page. The title is Mercury Decide (free). The price label is Free, and the page data lists a price of 0 USD. The page says output tokens are free and the model makes up to 14 decisions a second. The context line is 32,768 tokens. A search of that page for JevBench returned no match. openrouter.ai/inception/mercury-decide also returned HTTP 200, with the title Mercury Decide. That response’s HTML did not include a price.
This site's reading
Editorial notes evaluating claims against primary sources, contextualizing findings alongside related implementations, and defining technical terms.
Verify
Primary post is @_inception_ai, September 30, 2026, 19:12 UTC. The text calls Mercury Decide the most intelligent decision model on OpenRouter, citing JevBench v1.4, at up to 14 decisions per second. The attachment is a chess video. We did not transcribe a score from it. The free early-access link is https://openrouter.ai/inception/mercury-decide:free. That URL returned Not Found, and so did https://openrouter.ai/inception/mercury-decide.
OpenRouter's post, 19:08 UTC the same day, repeats the claim. The Mercury 2.5 page we opened describes Mercury Decide as a System One endpoint: a state plus typed questions, a choice, a score, or a yes/no, output tokens free, up to 14 decisions per second, the same /v1/systemone schema as Jev. The page's word for the probability is calibrated. We did not measure that.
On October 1, 2026 we opened https://benchmarkheaven.com/jev-models. The page still said v1.5.4 and 106 ranked systems. A search of that text for "Mercury Decide" returned no match. v1.4 on this desk is the older harmonic mean, where Jev is 63.29. The post does not print a Mercury Decide cell on that board or on v1.5. We did not rerun it.
Opened October 8, 2026, the free URL returned the model page. The title is Mercury Decide (free). The price label is Free. The page says output tokens are free and the model makes up to 14 decisions a second. The context line is 32,768 tokens. A search of that page for JevBench returned no match. The URL without :free also returned HTTP 200, with the title Mercury Decide. That response's HTML did not include a price.
Compare
Jev 1.13.0 on the v1.5.4 page is 72.1, median 0.62 seconds, about $0.032 per 1,000 decisions. The Inception post names JevBench v1.4 and does not print a number next to Mercury Decide. The 14 decisions a second is the post's rate and the Mercury 2.5 page's rate. It is not a latency we timed.
Solar Decide, a different System One endpoint on OpenRouter, prints 87.0% at p50 0.14 seconds against Jev at 86.1% and 0.30 seconds. That row has a denominator in its own story. This one does not. The free Mercury Decide URL returned Not Found, so this page does not list a working model id.
Opened October 8, 2026, the free URL returned the model page. The model id on that page is inception/mercury-decide:free. The price label is Free. The page prints no JevBench score.
Terms
- 14 decisions per second
- The rate in Inception's September 30 post and on OpenRouter's Mercury 2.5 page. The post does not print a JevBench score beside it.
- JevBench v1.4
- The edition the post cites. On this desk the v1.4.2.2 harmonic mean has Jev at 63.29. The live page on October 1 is v1.5.4, and a search of that text found no Mercury Decide.