News-Tracker

Modeller

AI & LLM · daglig briefing

Ændringer siden forrige snapshot

Score ændret

  • Granite 4.2 30B: intelligence 14.8 → 12.7 (-2.1).

Ny i HF-trending

  • perplexity-ai/pplx-decider-v1.1-27b er ny i Hugging Faces trending (score 74).
  • Phocinae/Phocinae-Largha-150M-v1 er ny i Hugging Faces trending (score 67).
  • moonshotai/Kimi-K3 er ny i Hugging Faces trending (score 51).
  • Cactus-Compute/needle3 er ny i Hugging Faces trending (score 48).

Intelligence Index — Artificial Analysis

#ModelIntel.Agentic CreatorUdgivet
1 Claude Opus 5.5 (Max, Default Fallback) 5 effort-varianter 57.6 Anthropic 2026-09-22
2 Claude Sonnet 5.5 (Max, Default Fallback) 5 effort-varianter 56 Anthropic 2026-09-28
3 Claude Fable 5.1 (Max, Default Fallback) 5 effort-varianter 53.4 57.9 Anthropic 2026-09-01
4 GPT-6 Astra (Max) 5 effort-varianter 52.7 51 OpenAI 2026-09-03
5 Gemini 4 Argon (High) 52.6 Google 2026-09-30
6 GPT-6.1 Sol (Max) 5 effort-varianter 51.8 OpenAI 2026-09-29
7 Claude Opus 5 (Max) 5 effort-varianter 50.8 56.5 Anthropic 2026-07-24
8 Claude Fable 5 (Max, Opus 4.8 Fallback) 49.6 50.7 Anthropic 2026-06-09
9 Muse Spark 1.3 (Max) 2 effort-varianter 48.1 55.5 Meta 2026-09-02
10 GPT-6 Sol (Max) 7 effort-varianter 47.6 OpenAI 2026-09-22
11 GPT-5.6 Sol (Max) 6 effort-varianter 47 50.2 OpenAI 2026-07-09
12 Grok 4.7 (Xhigh) 3 effort-varianter 46.4 SpaceXAI 2026-09-21
13 MiMo-V2.6-Pro 46.3 Xiaomi 2026-09-21
14 Qwen3.8 Max (0902) 2 effort-varianter 45.4 56 Alibaba 2026-09-02
15 GLM-5.3 (Max) 2 effort-varianter ⬇ hf.co/zai-org/GLM-5.3 44.8 53.1 Z AI 2026-08-18
16 Grok 4.6 (High) 4 effort-varianter 44.3 53 SpaceXAI 2026-08-12
17 Step 5 Preview ⬇ hf.co/stepfun-ai/Step-5-Preview 43.7 StepFun 2026-09-18
18 Kimi K3 (Max) 2 effort-varianter ⬇ hf.co/moonshotai/Kimi-K3 43.6 50 Kimi 2026-07-16
19 Claude Haiku 5.5 (Max) 5 effort-varianter 43.4 Anthropic 2026-10-07
20 GPT-5.6 Terra (Max) 6 effort-varianter 42.1 43.2 OpenAI 2026-07-09
21 Claude Opus 4.8 (Max) 41.8 41.9 Anthropic 2026-05-28
22 GLM-5.3-Flash ⬇ hf.co/zai-org/GLM-5.3-Flash 41.8 50.9 Z AI 2026-08-26
23 Ling 3.1 Flash ⬇ hf.co/inclusionAI/Ling-3.1-flash 41.1 InclusionAI 2026-10-01
24 Gemini 3.8 Flash (High) 3 effort-varianter 40.9 40.2 Google 2026-09-02
25 Claude Opus 4.7 (Max) 2 effort-varianter 40.7 38.6 Anthropic 2026-04-16
26 Qwen3.8 2.4T A95B ⬇ hf.co/Qwen/Qwen3.8-2.4T-A95B 39.9 50.1 Alibaba 2026-08-12
27 Qwen3.8-Flash-Next ⬇ hf.co/Qwen/Qwen3.8-Flash-Next 39.8 53.6 Alibaba 2026-08-26
28 Gemini 3.7 Flash (Medium) 3 effort-varianter 39.6 Google 2026-08-13
29 Muse Spark 1.2 (Xhigh) 39.6 43.2 Meta 2026-08-05
30 DeepSeek V4.1 Flash (Max) 2 effort-varianter ⬇ hf.co/deepseek-ai/DeepSeek-V4.1-Flash 39.5 DeepSeek 2026-09-10
31 GPT-5.4 (Xhigh) 3 effort-varianter 39 OpenAI 2026-03-05
32 Grok 4.5 (High) 38.8 41.2 SpaceXAI 2026-07-08
33 GPT-5.5 (Xhigh) 5 effort-varianter 38.4 36.4 OpenAI 2026-04-23
34 Mistral Large 4 Preview 38.4 Mistral 2026-10-06
35 Claude Sonnet 5 (Max) 6 effort-varianter 38.2 43.6 Anthropic 2026-06-30
36 GPT-6 Luna (Max) 6 effort-varianter 38.1 OpenAI 2026-09-22
37 MiMo-V2.6-Flash 37.9 Xiaomi 2026-09-21
38 GPT-5.6 Luna (Max) 6 effort-varianter 37.3 42.1 OpenAI 2026-07-09
39 DeepSeek V4 Pro 0813 (Max) 2 effort-varianter ⬇ hf.co/deepseek-ai/DeepSeek-V4-Pro 36 41.3 DeepSeek 2026-08-13
40 DeepSeek V4 Flash Vision (Max) 34.8 47.5 DeepSeek 2026-08-21

Hugging Face — trending familier

#Rod-repoTrendingLikes DownloadsParamsAlder
1 google/gemma-4-26B-A4B +2 derivater 1632 1901 903866 25.8B 7 d
2 Cloudflare/clef 1313 1897 10874 27.4B 8 d
3 google/embeddinggemma-2 +1 derivater 1182 1218 21148 744.4M 24 d
4 XingChen-AGI/Xing4.0-29B-A4B +1 derivater 617 669 36481 21 d
5 jialinyyzz/humanizer 599 640 23439 12.0B 28 d
6 Cloudflare/clef-flash +1 derivater 494 696 17587 9.4B 8 d
7 Lightricks/LTX-2.5 +2 derivater 494 6963 1688807 78 d
8 convaiinnovations/laya 375 5397 36328 421.3M 21 d
9 Qwen/Qwen3.8-27B +12 derivater 342 17297 6841660 27.8B 65 d
10 canberkkkkkk/ema-lightning 279 304 9467 7 d
11 Qwen/Qwen-Image-2.1 +7 derivater 219 3129 116957 7.1B 25 d
12 LiquidAI/d1-3B 199 201 5370 3.1B 4 d
13 Cactus-Compute/whistle 191 196 2594 8 d
14 Qwen/Qwen3.8-Flash-Next +6 derivater 181 6047 1640938 180.0B 46 d
15 deepseek-ai/DeepSeek-V4.1-Flash 153 4266 1282524 763.2B 29 d

Params er safetensors.total, altså lagrede vægte inklusive kvantiserede tensorer — ikke et mål for hvad modellen kræver af hardware.