Official model page not yet on record
Official pricing ↗Ranked 27 of the tracked routed-model set.
The page leads with what has been observed, then separates price, capability and history so a buyer can see both the signal and its limits.
Movement needs a second observation before it can be reported.
Per one million tokens. Cached input is not publicly disclosed.
Reasoning, tool calling and structured output are supported.
Live per-token pricing via OpenRouter, verified 18 Sep 2026. Batch and long-context tiers are not published in the feed, so they are omitted rather than estimated.
No cached-input price is published for this model, so cached tokens are costed at the full input rate. The cached-input amount is therefore an upper-bound estimate, not an observed discount.
An estimate from published per-token pricing and your assumptions. Actual costs vary by provider and usage tier.
Same workload as the calculator above, at each model's current price. Monthly USD.
Routed-token share over time, with no interpolation presented as evidence. A price line joins it as real changes are recorded.
Low cost · adoption flat. Observed relationship, not proof of causation.
Verified support, explicit non-support and missing disclosure read differently on purpose. A blank is never made to look like a negative finding.
| Context window | 1M |
| Reasoning | ✓ Supported |
| Tool calling | ✓ Supported |
| Structured output | ✓ Supported |
| Image input | — Not supported |
| Audio input | — Not supported |
| Video input | — Not supported |
| Open weights | ✓ Yes |
| API availability | ✓ Yes |
| Commercial license | Verify terms with the provider |
Median provider uptime across OpenRouter endpoints, last 30 days. Latency and throughput are not published by the source and are not estimated here.
In the week ending 18 Sep 2026, DeepSeek V4 Flash (free) entered the ranking this week with 0.58% of routed tokens and a peak rank of #27.
A sourced weekly interpretation of what is driving this model's routing is added as the record accumulates. yellow3 does not publish a reason for movement unless the routed-traffic data or a cited source supports it.
Dated developments concerning this model and its provider. Facts, evidence status and yellow3 analysis remain separate.
Routed share reached 0.58% in the seven days ending 18 Sep 2026. The record does not assign a cause.
ConcernsModel · DeepSeek V4 Flash (free)Open signal Close signalNo interpretation is published for this signal. The observation stands on its own.
OpenRouter usage represents routed developer traffic, not the whole AI market. Timing near an adoption change does not establish causation. Sources checked for this record: OpenRouter routed-traffic rankings, OpenRouter listed price. Provider announcements and independent reporting are not yet monitored and are absent rather than nil.
The current position stays connected to every observed week, price change and milestone.
| Week ending | Rank | Routed share | Δ share | Movement | Region rank | Status |
|---|---|---|---|---|---|---|
| 18 Sep 2026 | #27 | 0.58% | – | NEW | #16 | NEW |
| Date | Input / 1M | Output / 1M | Change | Source |
|---|---|---|---|---|
| 18 Sep 2026 | $0.00 | $0.00 | — | OpenRouter listed price |
This model is listed on more than one OpenRouter endpoint. Price observations recorded before 17 September 2026 did not identify the endpoint, so they could not be attributed reliably and have been withdrawn. Price history for this model includes only observations recorded from 17 September 2026, when yellow3 began recording the endpoint with each observation. See the correction record.
Region reflects where the model's developer is headquartered. Figures are OpenRouter routing traffic, aggregated over a trailing seven days - developer routing behaviour, not the whole market.
The public record shows today's evidence. yellow3 Model Intelligence adds longer history, watchlists, alerts, comparisons and decision reporting.