What changed in AI
Every material change the connectors observe becomes an event with a source. Dates are when the change occurred (effective date when known); hover a date for the observation time. Older events load with a cursor.
1,181 events since 20 May 2026 · ordered by occurrence
16 Sept 2026 16
GLM 5.3 Flash: supported parameters changed from ["frequency_penalty", "include_reasoning", "logit_bias", … to ["frequency_penalty", "include_reasoning", "logit_bias", …
Supported parametersfrequency_penalty, include_reasoning, logit_bias, logprobs, max_tokens, min_p, presence_penalty, reasoning, reasoning_effort, repetition_penalty, response_format, seed, stop, structured_outputs, temperature, tool_choice, tools, top_k, top_logprobs, top_p→frequency_penalty, include_reasoning, logit_bias, logprobs, max_tokens, min_p, parallel_tool_calls, presence_penalty, reasoning, reasoning_effort, repetition_penalty, response_format, seed, stop, structured_outputs, temperature, tool_choice, tools, top_k, top_logprobs, top_popenrouterNew artifact: unsloth/gemma-4-12b-it-GGUF (Unsloth)
huggingfaceNew artifact: mlx-community/GLM-OCR-4bit (MLX Community)
huggingfaceNew model: internlm2_5-7b (InternLM (Shanghai AI Laboratory))
huggingfaceNew model: Seed-OSS-36B-Base-woSyn (ByteDance)
huggingfaceUnslopNemo 12B: tool calling changed from true to false
Tool callingYes→NoopenrouterUnslopNemo 12B: supported parameters changed from ["frequency_penalty", "logit_bias", "logprobs", "max_toke… to ["frequency_penalty", "logit_bias", "logprobs", "max_toke…
Supported parametersfrequency_penalty, logit_bias, logprobs, max_tokens, presence_penalty, repetition_penalty, response_format, seed, stop, structured_outputs, temperature, tool_choice, tools, top_k, top_logprobs, top_p→frequency_penalty, logit_bias, logprobs, max_tokens, presence_penalty, repetition_penalty, seed, stop, structured_outputs, temperature, top_k, top_logprobs, top_popenrouterQwen3 VL 30B A3B Instruct: max output tokens changed from 32768 to 16384
Max output32.8K tokens→16.4K tokensopenrouterQwen3 VL 30B A3B Instruct: max output tokens changed from 16384 to 32768
Max output16.4K tokens→32.8K tokensopenrouterDeepSeek V3.2: parameter count changed from 685396921376 to 685355329792
Parameters685.4B→685.4BhuggingfaceNemotron 3 Ultra: max output tokens changed from 182520 to 32768
Max output182.5K tokens→32.8K tokensopenrouter
15 Sept 2026 27
gpt-5.5: reasoning changed from false to true
ReasoningNo→YesopenrouterQwen3.8 27B: reasoning changed from false to true
ReasoningNo→Yesopenrouterolmo-3-1-32b-instruct: reasoning changed from true to false
ReasoningYes→Noartificial_analysisQwen 3.8 Max: aa context window changed from 1000000 to 983616
Aa context window1,000,000→983,616artificial_analysisQwen3.8 Flash: supported parameters changed from ["frequency_penalty", "include_reasoning", "logit_bias", … to ["frequency_penalty", "include_reasoning", "logprobs", "m…
Supported parametersfrequency_penalty, include_reasoning, logit_bias, logprobs, max_tokens, min_p, presence_penalty, reasoning, repetition_penalty, response_format, seed, stop, structured_outputs, temperature, tool_choice, tools, top_k, top_logprobs, top_p→frequency_penalty, include_reasoning, logprobs, max_tokens, presence_penalty, reasoning, response_format, seed, stop, structured_outputs, temperature, tool_choice, tools, top_k, top_logprobs, top_popenrouterNew model: Gemini 3.8 Flash Cyber (Google DeepMind)
googleNew model: Gemini 3.8 Live Extended Thinking (Google)
googlepii-masking-vllm: last modified changed from 2026-09-13T22:30:27+00:00 to 2026-09-15T04:22:52+00:00
Last modified2026-09-13T22:30:27+00:00→2026-09-15T04:22:52+00:00huggingfaceMistral Small 4: context length changed from 262144 to 256000
Context window262.1K tokens→256K tokensmistralZ.ai GLM 5.2: context length changed from 1048576 to 1000000
Context window1.05M tokens→1M tokensmistralMistral 7B: version changed from 0.3 to 0.2
Version0.3→0.2mistralMistral 7B: official url changed from https://docs.mistral.ai/models/mistral-7b-0-3 to https://docs.mistral.ai/models/mistral-7b-0-2
Official pagehttps://docs.mistral.ai/models/mistral-7b-0-3→https://docs.mistral.ai/models/mistral-7b-0-2mistralMistral 7B: version changed from 0.2 to 0.3
Version0.2→0.3mistralMistral 7B: official url changed from https://docs.mistral.ai/models/mistral-7b-0-2 to https://docs.mistral.ai/models/mistral-7b-0-3
Official pagehttps://docs.mistral.ai/models/mistral-7b-0-2→https://docs.mistral.ai/models/mistral-7b-0-3mistralVoxtral Mini Transcribe: official url changed from https://docs.mistral.ai/models/voxtral-mini-transcribe-25-07 to https://docs.mistral.ai/models/voxtral-mini-25-07
Official pagehttps://docs.mistral.ai/models/voxtral-mini-transcribe-25-07→https://docs.mistral.ai/models/voxtral-mini-25-07mistralVoxtral Mini Transcribe: official url changed from https://docs.mistral.ai/models/voxtral-mini-25-07 to https://docs.mistral.ai/models/voxtral-mini-transcribe-25-07
Official pagehttps://docs.mistral.ai/models/voxtral-mini-25-07→https://docs.mistral.ai/models/voxtral-mini-transcribe-25-07mistralQwen3 Next 80B A3B Thinking: max output tokens changed from 235929 to 32768
Max output235.9K tokens→32.8K tokensopenrouterQwen3.5-35B-A3B: max output tokens changed from 16384 to 65536
Max output16.4K tokens→65.5K tokensopenrouterZ.ai GLM 5.2: max output tokens changed from 131072 to 128000
Max output131.1K tokens→128K tokensopenrouter
14 Sept 2026 7
pplx-pii-masking-vllm: release date changed from 2026-09-13 to 2026-08-02
Release date13 Sept 2026→2 Aug 2026huggingfacepplx-pii-masking-vllm: license changed from MIT to Other
LicenseMIT→Otherhuggingfacepplx-pii-masking-vllm: license raw changed from mit to other
License rawmit→otherhuggingfacepplx-pii-masking-vllm: last modified changed from 2026-09-13T22:30:27+00:00 to 2026-08-02T23:54:58+00:00
Last modified2026-09-13T22:30:27+00:00→2026-08-02T23:54:58+00:00huggingfaceHermes 3 70B Instruct: supported parameters changed from ["frequency_penalty", "logit_bias", "max_tokens", "min_p"… to ["frequency_penalty", "logit_bias", "max_tokens", "min_p"…
Supported parametersfrequency_penalty, logit_bias, max_tokens, min_p, presence_penalty, repetition_penalty, response_format, seed, stop, structured_outputs, temperature, top_k, top_p→frequency_penalty, logit_bias, max_tokens, min_p, presence_penalty, repetition_penalty, seed, stop, structured_outputs, temperature, top_k, top_popenrouterLlama 3.1 Euryale 70B v2.2: supported parameters changed from ["frequency_penalty", "logit_bias", "max_tokens", "min_p"… to ["frequency_penalty", "logit_bias", "max_tokens", "min_p"…
Supported parametersfrequency_penalty, logit_bias, max_tokens, min_p, presence_penalty, repetition_penalty, response_format, seed, stop, structured_outputs, temperature, tool_choice, tools, top_k, top_p→frequency_penalty, logit_bias, max_tokens, min_p, presence_penalty, repetition_penalty, seed, stop, structured_outputs, temperature, tool_choice, tools, top_k, top_popenrouter
Feed keyed on occurred_at. Backfill = history imported when a source is first crawled (excluded — tick the box in the rail to include it). Times are UTC.