What changed

  1. Anthropic: Claude Sonnet 4context window 1000000 → 200000source

  2. ByteDance Seed: Seed 1.6 Flashstatus active → deprecatedsource

    What it means The OpenRouter row for bytedance-seed/seed-1.6-flash gained an expiration_date of 2026-11-11 in the snapshot fetched 2026-09-23T23:47:07Z (it was null in the 2026-09-21T12:20:00Z one, with pricing, context length and created unchanged), so the registry's expiration-date rule (absent -> active, within 365 days -> deprecated, in the past -> retired) derives deprecated — a genuine removal notice, set in the same fetch on all three of bytedance-seed/seed-1.6, bytedance-seed/seed-1.6-flash and bytedance-seed/seed-2.0-code and still held in the 2026-09-24T12:05:12Z snapshot. It matters as a migration cue rather than an outage: the row was still listed as of that 2026-09-24 snapshot, but OpenRouter's model schema defines expiration_date as the date after which the model may be removed, so listing through 2026-11-11 is not guaranteed and removal may follow that date; if the row is still present once the date is past, the same rule derives retired. Meanwhile bytedance-seed/seed-2.0-lite, bytedance-seed/seed-2.0-mini and bytedance-seed/seed-2-1-turbo carry a null expiration_date in that 2026-09-24 snapshot — though the feed declares none of them a successor.

  3. ByteDance Seed: Seed 1.6status active → deprecatedsource

    What it means The OpenRouter row for bytedance-seed/seed-1.6 gained an expiration_date of 2026-11-11 in the snapshot fetched 2026-09-23T23:47:07Z (it was null in the 2026-09-21T12:20:00Z one, with pricing, context length and created unchanged), so the registry's expiration-date rule (absent -> active, within 365 days -> deprecated, in the past -> retired) derives deprecated — a genuine removal notice, set in the same fetch on all three of bytedance-seed/seed-1.6, bytedance-seed/seed-1.6-flash and bytedance-seed/seed-2.0-code and still held in the 2026-09-24T12:05:12Z snapshot. It matters as a migration cue rather than an outage: the row was still listed as of that 2026-09-24 snapshot, but OpenRouter's model schema defines expiration_date as the date after which the model may be removed, so listing through 2026-11-11 is not guaranteed and removal may follow that date; if the row is still present once the date is past, the same rule derives retired. Meanwhile bytedance-seed/seed-2.0-lite, bytedance-seed/seed-2.0-mini and bytedance-seed/seed-2-1-turbo carry a null expiration_date in that 2026-09-24 snapshot — though the feed declares none of them a successor.

  4. ByteDance Seed: Seed-2.0-Codestatus active → deprecatedsource

    What it means The OpenRouter row for bytedance-seed/seed-2.0-code gained an expiration_date of 2026-11-11 in the snapshot fetched 2026-09-23T23:47:07Z (it was null in the 2026-09-21T12:20:00Z one, with pricing, context length and created unchanged), so the registry's expiration-date rule (absent -> active, within 365 days -> deprecated, in the past -> retired) derives deprecated — a genuine removal notice, set in the same fetch on all three of bytedance-seed/seed-1.6, bytedance-seed/seed-1.6-flash and bytedance-seed/seed-2.0-code and still held in the 2026-09-24T12:05:12Z snapshot. It matters as a migration cue rather than an outage: the row was still listed as of that 2026-09-24 snapshot, but OpenRouter's model schema defines expiration_date as the date after which the model may be removed, so listing through 2026-11-11 is not guaranteed and removal may follow that date; if the row is still present once the date is past, the same rule derives retired. Meanwhile bytedance-seed/seed-2.0-lite, bytedance-seed/seed-2.0-mini and bytedance-seed/seed-2-1-turbo carry a null expiration_date in that 2026-09-24 snapshot — though the feed declares none of them a successor.

  5. inclusionAI: Ling 3.0 Flash VLcontext window 131072 → 262144source

  6. Nex AGI: Nex-N2.5-Mini (free)status active → deprecatedsource

    What it means The free OpenRouter row nex-agi/nex-n2.5-mini:free gained an expiration_date of 2026-09-25 in the snapshot fetched 2026-09-23T23:47:07Z (null in the 2026-09-21T12:20:00Z one; prices still 0/0), so the registry's expiration-date rule derives deprecated now and retired once that date is past — a two-day notice, the same date nex-agi/nex-n2.5-pro:free carries, and the same short-notice pattern nex-agi/nex-n2-mini showed earlier on this source. It matters because no paid fallback survives in the feed: the paid nex-agi/nex-n2.5-mini row that arrived in the same fetch (change openrouter-models|3c38df3673431ec9|168e394c76befa88|nex-agi/nex-n2.5-mini|$arrival) was gone from the 2026-09-24T12:05:12Z snapshot (change openrouter-models|168e394c76befa88|fe2b09d80953c7be|nex-agi/nex-n2.5-mini|$retirement), so a reader relying on the free endpoint has no same-model row to move to.

Observed mechanically from public sources; every line carries the source it was read from. The last 30 days as a table · RSS

Impossible → Routine

Capabilities that were research results, and the date each became something anyone could buy. Both ends dated, both ends sourced.

A forecast that was learned, not simulated

Forecasting global weather with a learned model instead of simulated atmospheric physics.

Impossible

GraphCast, a research model, beats the industry gold-standard physics forecast on more than 90% of tested variables.

10-day forecast in under a minute on one TPU machine

source

2 years, 9 months

Routine

WeatherNext 3 becomes the new leader on Brightband's Operational WeatherBench, a live third-party comparison of AI and physics global medium-range models, edging out its predecessor WeatherNext 2.

lowest 2m-temperature error on 26 of the last 30 days in August

source

Real work on a real desktop

Completing multi-step tasks on a real computer — opening applications, clicking through interfaces, finishing the job.

Impossible

The OSWorld benchmark arrives and the best model completes 12.24% of its 369 real-computer tasks.

12.24% on OSWorld, the original 369-task set

source

2 years, 3 months

Routine

Qwen3.8-Max posts 86.1% on OSWorld-Verified, the repaired 2025 revision of the benchmark, with competing agents from two other labs already above 83%.

86.1% on OSWorld-Verified, the 2025 revision

source

All 27 dated pairs