400
changes shipped in the last two months that nobody announced. Prices moved, context windows shrank, capabilities disappeared — no email, no changelog entry, no error at runtime.
145
price changes
you pay more for the same call
293
capacity cuts
prompts truncate, no error
123
capabilities removed
a feature quietly went away
258
announced retirements
these ones you were told about
Evidence behind the numbers: 74 models confirmed against the provider's own API, 152 with dates taken from the provider's own documentation (extracted by a model, so labelled inferred), the remaining 2669 from third-party catalogs only. Every model page says which of the three it rests on.
Biggest silent changes
| Impact | Provider | What changed | Detected |
|---|---|---|---|
| truncation | gemini | -99% ctx · gemini/gemini-2.5-pro-preview-tts: context_tokens cut 1,048,576 to 8,192 | 2026-09-16 |
| truncation | bedrock | -98% max out · ai21.jamba-1-5-large-v1:0: max_output_tokens cut 256,000 to 4,096 +1 alias inferred | 2026-10-12 |
| truncation | openrouter | -98% max out · openrouter/moonshotai/kimi-k3:batch: max_output_tokens cut 943,718 to 16,384 | 2026-09-22 |
| truncation | bedrock_converse | -97% max out · zai.glm-4.7-flash: max_output_tokens cut 128,000 to 4,000 +1 alias | 2026-09-21 |
| truncation | thedrummer | -97% max out · thedrummer/unslopnemo-12b: max_output_tokens cut 1,024,000 to 32,768 inferred | 2026-08-24 |
| cost | z-ai | 28.85x price · z-ai/glm-5.3-flash: input_price_per_mtok $0.0052/Mtok to $0.15/Mtok | 2026-10-23 |
| truncation | openai | -97% max out · openai/gpt-4.1-nano: max_output_tokens cut 942,818 to 32,768 | 2026-08-30 |
| truncation | nvidia | -96% max out · nvidia/nemotron-3-ultra-550b-a55b: max_output_tokens cut 461,059 to 16,384 | 2026-08-28 |
| truncation | azure_ai | -96% max out · azure_ai/grok-4.3: max_output_tokens cut 200,000 to 8,192 | 2026-09-28 |
| cost | ~deepseek | 20.83x price · ~deepseek/deepseek-flash-latest: output_price_per_mtok $0.02/Mtok to $0.50/Mtok | 2026-10-20 |
| truncation | bedrock_converse | -95% max out · deepseek.v3.2: max_output_tokens cut 163,840 to 8,000 inferred | 2026-09-21 |
| truncation | bedrock | -95% max out · bedrock/ap-northeast-1/deepseek.v3.2: max_output_tokens cut 163,840 to 8,000 +7 aliases inferred | 2026-10-08 |
| cost | deepseek | 19.80x price · deepseek/deepseek-v4-flash-0731: cache_read_price_per_mtok $0.0014/Mtok to $0.03/Mtok | 2026-10-25 |
| truncation | bedrock_converse | -94% max out · moonshotai.kimi-k2.5: max_output_tokens cut 262,144 to 16,000 | 2026-09-21 |
| truncation | bedrock | -94% max out · bedrock/ap-northeast-1/moonshotai.kimi-k2-thinking: max_output_tokens cut 262,144 to 16,000 +15 aliases inferred | 2026-10-08 |
| truncation | qwen | -94% max out · qwen/qwen3-next-80b-a3b-instruct: max_output_tokens cut 262,144 to 16,384 +1 alias | 2026-08-19 |
| truncation | azure_ai | -94% max out · azure_ai/grok-4: max_output_tokens cut 131,072 to 8,192 +1 alias inferred | 2026-09-28 |
| truncation | bedrock_converse | -94% max out · qwen.qwen3-235b-a22b-2507-v1:0: max_output_tokens cut 131,072 to 8,192 | 2026-10-07 |
All 400 changes → · Check your own models →
Announced retirements
The part providers do tell you about. Included for completeness — it is the smaller half of the problem.
| Model | Vendor | Sunset | Remaining | Severity |
|---|---|---|---|---|
| arcee-ai/trinity-large-thinking | arcee-ai | T-1d | critical | |
| openai/gpt-realtime-2 | openai | T-6d | critical | |
| mistral/zai-glm-5-2 | mistral | T-6d | critical | |
| poolside/laguna-s-2.1 | poolside | T-6d | critical | |
| poolside/laguna-s-2.1:free | poolside | T-6d | critical | |
| poolside/laguna-xs-2.1 | poolside | T-6d | critical | |
| poolside/laguna-xs-2.1:free | poolside | T-6d | critical | |
| azure/sora-2 | azure | T-8d | critical | |
| nousresearch/hermes-3-llama-3.1-405b | nousresearch | T-8d | critical | |
| nousresearch/hermes-3-llama-3.1-70b | nousresearch | T-8d | critical |