llm-catalog-archive
RSS Repository

The archive

Everything

7,311 items derived by replay from git history over raw/, newest first. Every sentence names the artifact it was read out of and links those bytes at the commit that stored them.

The 12 most recently active threads, as tabs. The other 662 are in the threads index, uncapped. Where this browser can draw it, the same 12 tabs are a wall in three dimensions; the list is what they link, and the list is what is here when they are not.

What happened

Announcements, incidents, leaks, and models arriving or leaving a catalogue. The full stream below is chronological and includes these again, plus the routine catalogue activity they would otherwise be buried under.

  1. post_published The openai-news-feed published a post titled "Benefit Street Partners answers questions faster with Codex" at "https://openai.com/index/benefit-street-partners". origin
  2. model_added The catalog id "unbiased/pareto-26.10" entered OpenRouter's catalog. origin
  3. model_added The catalog id "deepseek/deepseek-v4.1-flash-1017" entered OpenRouter's catalog. origin
  4. model_added The catalog id "elorian/glimpse" entered OpenRouter's catalog. origin
  5. post_published The huggingface-blog-feed published a post titled "Welcome Beam: a 501B open model you can own, customize, and scale" at "https://huggingface.co/blog/reflection-beam". observed
  6. upstream_pr_opened A pull request numbered "vllm-project/vllm#62057" is titled "[ROCm][Model] Add Beam 1.0 support on ROCm " in the collected search payload. observed
  7. post_listed The anthropic-sitemap index listed a URL it had not listed before: "https://www.anthropic.com/research/ai-expectations-gap". observed
  8. post_listed The anthropic-sitemap index listed a URL it had not listed before: "https://www.anthropic.com/research/designing-efficient-quantum-circuits". observed
By lab

deepseek2,294 zai983 moonshot533 openai376 google163 nvidia114 qwen359 anthropic487 mistral409 meta79 together114 minimax55 xai52 perplexity85 cohere1

Showing 150 items of 7,311: 3,288 held back so that no single capture fills the page, and 3,873 beyond this page's limit of 150. Nothing is dropped: every item is also on its micro-category page, on its lab page where it has one, and on the entity thread it attaches to where one could be resolved, and none of those are capped.

2026-10-18

  1. a3a7757 openrouter-models origin
    10 listed values moved across 5 models in this capture. Both numbers are the artifact's own; the percentage is this page's arithmetic over them. Each model links to its thread, which carries the full sentence and every other item recorded against it.
    FieldFromToChange
    deepseek/deepseek-v4-flash-0731
    input_cache_read 0.0000000199 0.0000000151 24%
    prompt 0.0000000199 0.0000000151 24%
    tencent/hy3
    completion 0.000000528 0.00000033 37%
    input_cache_read 0.000000033 0.000000020625 37%
    prompt 0.000000132 0.0000000825 37%
    tencent/hy4-preview
    completion 0.000002501 0.0000022509 10%
    input_cache_read 0.000000042 0.0000000378 10%
    prompt 0.000000834 0.0000007506 10%
    z-ai/glm-5.3
    completion 0.0000025 0.0000018 28%
    ~deepseek/deepseek-v4-flash-latest
    prompt 0.00000001592 0.00000001208 24%
  2. 10bb771 openrouter-models origin
    12 listed values moved across 5 models in this capture. Both numbers are the artifact's own; the percentage is this page's arithmetic over them. Each model links to its thread, which carries the full sentence and every other item recorded against it.
    FieldFromToChange
    deepseek/deepseek-v4-flash-0731
    input_cache_read 0.0000000151 0.0000000199 32%
    prompt 0.0000000151 0.0000000199 32%
    deepseek/deepseek-v4.1-flash
    prompt 0.0000000648 0.00000004 38%
    ~deepseek/deepseek-v4-flash-latest
    completion 0.00000045 0.000000143616 68%
    input_cache_read 0.0000000151 0.000000001523 90%
    prompt 0.0000000151 0.00000001592 5%
    ~moonshotai/kimi-latest
    completion 0.000014 0.000008 43%
    input_cache_read 0.00000027 0.00000018 33%
    prompt 0.00000027 0.00000025 7%
    ~z-ai/glm-latest
    completion 0.0000044 0.0000022 50%
    input_cache_read 0.0000000588 0.00000007 19%
    prompt 0.0000000632 0.000000075 19%
  3. e0ea159 openrouter-models origin
    6 listed values moved across 3 models in this capture. Both numbers are the artifact's own; the percentage is this page's arithmetic over them. Each model links to its thread, which carries the full sentence and every other item recorded against it.
    FieldFromToChange
    deepseek/deepseek-v4-flash-0731
    completion 0.0000003 0.00000045 50%
    input_cache_read 0.00000003 0.0000000151 50%
    prompt 0.0000001 0.0000000151 85%
    z-ai/glm-5.3
    completion 0.0000048 0.0000025 48%
    ~deepseek/deepseek-v4-flash-latest
    input_cache_read 0.0000000086 0.0000000151 76%
    prompt 0.0000000086 0.0000000151 76%
  4. 553b2cb openrouter-models origin
    7 listed values moved across 4 models in this capture. Both numbers are the artifact's own; the percentage is this page's arithmetic over them. Each model links to its thread, which carries the full sentence and every other item recorded against it.
    FieldFromToChange
    deepseek/deepseek-v4.1-flash
    prompt 0.0000000552 0.0000000648 17%
    z-ai/glm-5.3
    completion 0.0000032 0.0000048 50%
    ~deepseek/deepseek-flash-latest
    completion 0.0000005 0.0000006 20%
    prompt 0.00000008 0.00000005 38%
    ~deepseek/deepseek-v4-flash-latest
    completion 0.000000143616 0.00000045 213%
    input_cache_read 0.000000001523 0.0000000086 465%
    prompt 0.00000000912 0.0000000086 6%
  5. d52eb64 openai-news-feed origin
    1. post_published

      The openai-news-feed published a post titled "Benefit Street Partners answers questions faster with Codex" at "https://openai.com/index/benefit-street-partners".

      origin · raw artifact at this commit

      api-surfaceopenai.com/index labopenai

  6. 4ffd586 openrouter-models origin
    12 listed values shown of 17 that moved across 9 models in this capture. Both numbers are the artifact's own; the percentage is this page's arithmetic over them. Each model links to its thread, which carries the full sentence and every other item recorded against it.
    FieldFromToChange
    deepseek/deepseek-v4-flash-0731
    completion 0.00000128 0.0000003 77%
    input_cache_read 0.0000000125 0.00000003 140%
    prompt 0.0000000125 0.0000001 700%
    deepseek/deepseek-v4.1-flash
    prompt 0.0000000568 0.0000000552 3%
    google/gemma-3-27b-it
    completion 0.00000045 0.00000035 22%
    nvidia/nemotron-3.5-super-vl
    completion 0.0000005 0.0000004 20%
    input_cache_read 0.00000002 0.00000008 300%
    prompt 0.00000015 0.0000001 33%
    qwen/qwen3-30b-a3b-instruct-2507
    prompt 0.0000001 0.00000009 10%
    tencent/hy3
    completion 0.00000033 0.000000528 60%
    input_cache_read 0.000000020625 0.000000033 60%
    prompt 0.0000000825 0.000000132 60%

    5 further items from this capture are not shown here. Every one of them is on its micro-category page, which is uncapped.

2026-10-17

  1. c032f44 openrouter-models origin
    1. model_added

      The catalog id "unbiased/pareto-26.10" entered OpenRouter's catalog.

      origin · raw artifact at this commit

      modelunbiased/pareto-26.10

      catalog created1792073620
      first-seen worst-case error348,766 seconds, measured from this source's capture history
      first seen in the catalognot shown: the worst-case error is wider than a day
    11 listed values shown of 18 that moved across 6 models in this capture. Both numbers are the artifact's own; the percentage is this page's arithmetic over them. Each model links to its thread, which carries the full sentence and every other item recorded against it.
    FieldFromToChange
    ~unbiased/pareto-latest
    context_length 262144 1048576 300%
    deepseek/deepseek-v4-flash-0731
    input_cache_read 0.0000000189 0.0000000125 34%
    prompt 0.0000000189 0.0000000125 34%
    deepseek/deepseek-v4.1-flash
    completion 0.0000006 0.000000144 76%
    input_cache_read 0.000000003 0.0000000016 47%
    prompt 0.00000015 0.0000000568 62%
    z-ai/glm-5.3-flash
    completion 0.0000005 0.0000002 60%
    input_cache_read 0.00000003 0.000000006 80%
    prompt 0.00000015 0.000000048 68%
    ~deepseek/deepseek-v4-flash-latest
    completion 0.00000128 0.000000143616 89%
    input_cache_read 0.0000000189 0.000000001523 92%

    7 further items from this capture are not shown here. Every one of them is on its micro-category page, which is uncapped.

  2. 100f2db openrouter-models origin
    12 listed values moved across 5 models in this capture. Both numbers are the artifact's own; the percentage is this page's arithmetic over them. Each model links to its thread, which carries the full sentence and every other item recorded against it.
    FieldFromToChange
    deepseek/deepseek-v4-flash-0731
    input_cache_read 0.0000000067 0.0000000189 182%
    prompt 0.0000000067 0.0000000189 182%
    nvidia/nemotron-3.5-super-vl
    completion 0.0000004 0.0000005 25%
    input_cache_read 0.00000008 0.00000002 75%
    prompt 0.0000001 0.00000015 50%
    z-ai/glm-4.7
    completion 0.0000022 0.00000198 10%
    input_cache_read 0.00000011 0.000000099 10%
    prompt 0.0000006 0.00000054 10%
    ~deepseek/deepseek-flash-latest
    completion 0.00000055 0.0000005 9%
    prompt 0.00000009 0.00000008 11%
    ~deepseek/deepseek-v4-flash-latest
    input_cache_read 0.0000000067 0.0000000189 182%
    prompt 0.0000000067 0.0000000189 182%
  3. 30da384 openrouter-models origin
    7 listed values moved across 3 models in this capture. Both numbers are the artifact's own; the percentage is this page's arithmetic over them. Each model links to its thread, which carries the full sentence and every other item recorded against it.
    FieldFromToChange
    deepseek/deepseek-v4-flash-0731
    input_cache_read 0.0000000125 0.0000000067 46%
    prompt 0.0000000125 0.0000000067 46%
    ~deepseek/deepseek-flash-latest
    completion 0.0000005 0.00000055 10%
    input_cache_read 0.000000002 0.000000008 300%
    prompt 0.0000001 0.00000009 10%
    ~deepseek/deepseek-v4-flash-latest
    input_cache_read 0.0000000125 0.0000000067 46%
    prompt 0.0000000125 0.0000000067 46%
  4. dc2fcea openrouter-models origin
    1. expiration_scheduled confirmed-artifact

      OpenRouter's catalog recorded an expiration_date of "2026-12-31" for "z-ai/glm-4.7".

      origin · raw artifact at this commit

      modelz-ai/glm-4.7 labzai

      catalog idz-ai/glm-4.7
      expiration_date2026-12-31
      sentinel checka year at or past 2090 is read as the vendor spelling "no expiry" and is not listed here
    2. expiration_set

      OpenRouter's catalog recorded an expiration_date of "2026-12-31" for "z-ai/glm-4.7".

      origin · raw artifact at this commit

      modelz-ai/glm-4.7 labzai

    10 listed values shown of 13 that moved across 5 models in this capture. Both numbers are the artifact's own; the percentage is this page's arithmetic over them. Each model links to its thread, which carries the full sentence and every other item recorded against it.
    FieldFromToChange
    deepseek/deepseek-v4-flash-0731
    input_cache_read 0.0000000052 0.0000000125 140%
    prompt 0.0000000052 0.0000000125 140%
    tencent/hy3
    completion 0.000000528 0.00000033 37%
    input_cache_read 0.000000033 0.000000020625 37%
    prompt 0.000000132 0.0000000825 37%
    tencent/hy4-preview
    completion 0.000002501 0.0000022509 10%
    input_cache_read 0.000000042 0.0000000378 10%
    prompt 0.000000834 0.0000007506 10%
    z-ai/glm-4.7
    completion 0.00000198 0.0000022 11%
    input_cache_read 0.000000099 0.00000011 11%

    3 further items from this capture are not shown here. Every one of them is on its micro-category page, which is uncapped.

  5. 7476225 openrouter-models origin
    12 listed values shown of 23 that moved across 8 models in this capture. Both numbers are the artifact's own; the percentage is this page's arithmetic over them. Each model links to its thread, which carries the full sentence and every other item recorded against it.
    FieldFromToChange
    deepseek/deepseek-v4-flash-0731
    completion 0.0000002971 0.00000128 331%
    input_cache_read 0.0000000102 0.0000000052 49%
    prompt 0.0000000102 0.0000000052 49%
    z-ai/glm-5.3-flash
    completion 0.0000004182 0.0000005 20%
    input_cache_read 0.0000000099 0.00000003 203%
    prompt 0.0000000308 0.00000015 387%
    ~deepseek/deepseek-flash-latest
    completion 0.0000006 0.0000005 17%
    input_cache_read 0.000000003 0.000000002 33%
    prompt 0.00000015 0.0000001 33%
    ~deepseek/deepseek-pro-latest
    completion 0.00000039996 0.000008 1900%
    input_cache_read 0.000000004242 0.00000012 2729%
    prompt 0.00000013332 0.00000012 10%

    11 further items from this capture are not shown here. Every one of them is on its micro-category page, which is uncapped.

  6. b9174a4 openrouter-models origin
    1. model_added

      The catalog id "deepseek/deepseek-v4.1-flash-1017" entered OpenRouter's catalog.

      origin · raw artifact at this commit

      modeldeepseek/deepseek-v4.1-flash-1017 labdeepseek

      catalog created1792225624
      first-seen worst-case error348,766 seconds, measured from this source's capture history
      first seen in the catalognot shown: the worst-case error is wider than a day
    11 listed values shown of 27 that moved across 9 models in this capture. Both numbers are the artifact's own; the percentage is this page's arithmetic over them. Each model links to its thread, which carries the full sentence and every other item recorded against it.
    FieldFromToChange
    deepseek/deepseek-v4-flash-0731
    completion 0.00000014 0.0000002971 112%
    input_cache_read 0.000000014 0.0000000102 27%
    prompt 0.00000007 0.0000000102 85%
    deepseek/deepseek-v4.1-flash
    completion 0.00000055 0.0000006 9%
    input_cache_read 0.000000009 0.000000003 67%
    prompt 0.00000004 0.00000015 275%
    z-ai/glm-5.3-flash
    completion 0.0000005 0.0000004182 16%
    input_cache_read 0.00000003 0.0000000099 67%
    prompt 0.00000015 0.0000000308 79%
    ~deepseek/deepseek-flash-latest
    completion 0.00000055 0.0000006 9%
    input_cache_read 0.000000009 0.000000003 67%

    16 further items from this capture are not shown here. Every one of them is on its micro-category page, which is uncapped.

  7. 46c9e41 openrouter-models origin
    1. model_added

      The catalog id "elorian/glimpse" entered OpenRouter's catalog.

      origin · raw artifact at this commit

      modelelorian/glimpse

      catalog created1792206629
      first-seen worst-case error348,766 seconds, measured from this source's capture history
      first seen in the catalognot shown: the worst-case error is wider than a day
    11 listed values shown of 30 that moved across 10 models in this capture. Both numbers are the artifact's own; the percentage is this page's arithmetic over them. Each model links to its thread, which carries the full sentence and every other item recorded against it.
    FieldFromToChange
    deepseek/deepseek-v3.2
    completion 0.0000008 0.00000042 48%
    input_cache_read 0.00000014 0.000000028 80%
    prompt 0.000000259 0.00000028 8%
    deepseek/deepseek-v4-flash-0731
    completion 0.00000128 0.00000014 89%
    input_cache_read 0.0000000047 0.000000014 198%
    prompt 0.0000000047 0.00000007 1389%
    deepseek/deepseek-v4-pro
    completion 0.0000004524 0.0000004176 8%
    input_cache_read 0.00000001885 0.0000000174 8%
    prompt 0.0000002262 0.0000002088 8%
    deepseek/deepseek-v4.1-flash
    completion 0.000000144 0.00000055 282%
    input_cache_read 0.0000000016 0.000000009 462%

    19 further items from this capture are not shown here. Every one of them is on its micro-category page, which is uncapped.

  8. 7641a9f huggingface-blog-feed observed
    1. post_published

      The huggingface-blog-feed published a post titled "Welcome Beam: a 501B open model you can own, customize, and scale" at "https://huggingface.co/blog/reflection-beam".

      observed · raw artifact at this commit

      api-surfacehuggingface.co/blog

  9. 9b6be30 vllm-pulls observed
    1. upstream_pr_opened confirmed-artifact

      A pull request numbered "vllm-project/vllm#62057" is titled "[ROCm][Model] Add Beam 1.0 support on ROCm " in the collected search payload.

      observed · raw artifact at this commit

      held: no entity was mechanically extractable

      repositoryvllm-project/vllm
      pull request62057
      title[ROCm][Model] Add Beam 1.0 support on ROCm
      architecture named in the titleBeam 1.0
      pull request URLhttps://github.com/vllm-project/vllm/pull/62057
      created_at2026-10-16T18:04:50Z
      stateopen

      Check for a published weights repository under this name: https://huggingface.co/api/models?search=Beam%201.0. An empty array is a statement about that search, not about what exists.

  10. 9ab813c anthropic-sitemap observed
    1. post_listed

      The anthropic-sitemap index listed a URL it had not listed before: "https://www.anthropic.com/research/ai-expectations-gap".

      observed · raw artifact at this commit

      api-surfacewww.anthropic.com/research labanthropic

    2. post_listed

      The anthropic-sitemap index listed a URL it had not listed before: "https://www.anthropic.com/research/designing-efficient-quantum-circuits".

      observed · raw artifact at this commit

      api-surfacewww.anthropic.com/research labanthropic

    3. post_listed

      The anthropic-sitemap index listed a URL it had not listed before: "https://www.anthropic.com/research/ecdsa-curve-origins".

      observed · raw artifact at this commit

      api-surfacewww.anthropic.com/research labanthropic

    4. post_listed

      The anthropic-sitemap index listed a URL it had not listed before: "https://www.anthropic.com/research/employment-at-anthropic".

      observed · raw artifact at this commit

      api-surfacewww.anthropic.com/research labanthropic

    5. post_listed

      The anthropic-sitemap index listed a URL it had not listed before: "https://www.anthropic.com/research/expectations-for-the-future".

      observed · raw artifact at this commit

      api-surfacewww.anthropic.com/research labanthropic

  11. f920d3d openai-news-feed origin
    1. post_published

      The openai-news-feed published a post titled "How Itaú is building generative AI for its entire customer base" at "https://openai.com/index/itau-generative-ai".

      origin · raw artifact at this commit

      api-surfaceopenai.com/index labopenai

    2. post_published

      The openai-news-feed published a post titled "Stack Overflow for Agents triples signups with ChatGPT plugin" at "https://openai.com/index/stack-overflow-chatgpt-plugin".

      origin · raw artifact at this commit

      api-surfaceopenai.com/index labopenai

  12. 6cf9ecf claude-llms-txt origin
    1. doc_added

      The claude-llms-txt index added an entry titled "Claude Platform on AWS" at "https://platform.claude.com/docs/en/manage-claude/cmek-claude-platform-on-aws.md".

      origin · raw artifact at this commit

      api-surfaceplatform.claude.com/docs/en/manage-claude labanthropic

  13. 8267cfc openai-llms-txt origin
    1. doc_added

      The openai-llms-txt index added an entry titled "Browser and desktop drivers" at "https://developers.openai.com/api/docs/guides/tools-computer-use-sdk-drivers.md".

      origin · raw artifact at this commit

      api-surfacedevelopers.openai.com/api/docs/guides labopenai

  14. 6dd064c openrouter-models origin
    1. model_added

      The catalog id "qwen/qwen3.7-plus:batch" entered OpenRouter's catalog.

      origin · raw artifact at this commit

      modelqwen/qwen3.7-plus:batch labqwen

      catalog created1780491783
      first-seen worst-case error348,766 seconds, measured from this source's capture history
      first seen in the catalognot shown: the worst-case error is wider than a day
    11 listed values shown of 29 that moved across 10 models in this capture. Both numbers are the artifact's own; the percentage is this page's arithmetic over them. Each model links to its thread, which carries the full sentence and every other item recorded against it.
    FieldFromToChange
    deepseek/deepseek-v4-flash-0731
    input_cache_read 0.0000000049 0.0000000047 4%
    prompt 0.0000000049 0.0000000047 4%
    deepseek/deepseek-v4-pro
    completion 0.000000553668 0.0000004524 18%
    input_cache_read 0.0000000230695 0.00000001885 18%
    prompt 0.000000276834 0.0000002262 18%
    deepseek/deepseek-v4.1-flash
    completion 0.0000006435 0.000000144 78%
    input_cache_read 0.0000000054 0.0000000016 70%
    prompt 0.000000036 0.000000032 11%
    nvidia/nemotron-3.5-super-vl
    completion 0.0000005 0.0000004 20%
    input_cache_read 0.00000002 0.00000008 300%
    prompt 0.00000015 0.0000001 33%

    18 further items from this capture are not shown here. Every one of them is on its micro-category page, which is uncapped.