News-Tracker

Modeller

AI & LLM · daglig briefing

Ændringer siden forrige snapshot

Ny i indekset

  • Ling 3.1 Flash er ny i Artificial Analysis' indeks (uden score, InclusionAI).

Ny i HF-trending

  • higurashi-ai/Haru-4B er ny i Hugging Faces trending (score 155).
  • LiquidAI/Liquid-Codec er ny i Hugging Faces trending (score 79).

Intelligence Index — Artificial Analysis

#ModelIntel.Agentic CreatorUdgivet
1 Claude Fable 5.2 (Max, Default Fallback) 5 effort-varianter 64.6 Anthropic 2026-11-09
2 Gemini 4 Carbon (High) 3 effort-varianter 60 Google 2026-10-28
3 Claude Opus 5.5 (Max, Default Fallback) 5 effort-varianter 59.1 Anthropic 2026-09-22
4 GPT-6 Astra (Max) 5 effort-varianter 56.9 OpenAI 2026-09-03
5 GPT-6.1 Sol (Max) 5 effort-varianter 54.6 OpenAI 2026-09-29
6 Claude Sonnet 5.5 (Max, Default Fallback) 5 effort-varianter 54.5 Anthropic 2026-09-28
7 Gemini 4 Hydrogen (Xhigh) 4 effort-varianter 53.9 Google 2026-11-12
8 GPT-6.1 Terra (Max) 6 effort-varianter 51.3 OpenAI 2026-10-26
9 Gemini 4 Argon (High) 2 effort-varianter 50.4 Google 2026-09-30
10 Muse Wonder Full (Max) 50.1 Meta 2026-10-12
11 Claude Fable 5.1 (Xhigh, Default Fallback) 5 effort-varianter 50 Anthropic 2026-09-01
12 Kimi K3.1 Pro ⬇ hf.co/moonshotai/Kimi-K3.1-Pro 46.5 Kimi 2026-10-23
13 Claude Opus 5 (Max) 5 effort-varianter 44.2 Anthropic 2026-07-24
14 GPT-6 Sol (Max) 7 effort-varianter 44.2 OpenAI 2026-09-22
15 Muse Wonder (Max) 2 effort-varianter ⬇ hf.co/meta-01/muse-wonder 43.6 Meta 2026-10-12
16 Muse Spark 1.3 (Max) 2 effort-varianter 36.6 Meta 2026-09-02
17 Claude Haiku 5.5 (Max) 5 effort-varianter 36.5 Anthropic 2026-10-07
18 Muse Ember (Max) 2 effort-varianter 36 Meta 2026-11-09
19 Grok 4.7 (Xhigh) 3 effort-varianter 35.5 SpaceXAI 2026-09-21
20 Tencent Hy4 34.6 Tencent 2026-11-03
21 Qwen3.8 Max (0902) 2 effort-varianter 33.7 Alibaba 2026-09-02
22 DeepSeek V4.1 Flash 1017 (Max) 3 effort-varianter ⬇ hf.co/deepseek-ai/DeepSeek-V4.1-Flash-1017 32.6 DeepSeek 2026-10-17
23 GPT-5.6 Terra (Max) 6 effort-varianter 32.3 OpenAI 2026-07-09
24 GLM-5.3 (Max) 2 effort-varianter ⬇ hf.co/zai-org/GLM-5.3 31.7 Z AI 2026-08-18
25 Grok 4.6 (Xhigh) 4 effort-varianter 31 SpaceXAI 2026-08-12
26 Gemini 3.8 Flash (High) 3 effort-varianter 30.8 Google 2026-09-02
27 Inkling 2 Small ⬇ hf.co/thinkingmachines/Inkling-2-Small 30.6 Thinking Machines 2026-11-05
28 MiMo-V2.6-Pro 29.9 Xiaomi 2026-09-21
29 Ling 3.1 Flash 1019 ⬇ hf.co/inclusionAI/Ling-3.1-flash 29.6 InclusionAI 2026-10-22
30 MiniMax-M3.1-Flash ⬇ hf.co/MiniMaxAI/MiniMax-M3.1-Flash 29.1 MiniMax 2026-10-19
31 Qwen3.8-Flash-Next ⬇ hf.co/Qwen/Qwen3.8-Flash-Next 28.6 Alibaba 2026-08-26
32 Step 5 Preview ⬇ hf.co/stepfun-ai/Step-5-Preview 28.5 StepFun 2026-09-18
33 GPT-6 Luna (Max) 6 effort-varianter 27.9 OpenAI 2026-09-22
34 GLM-5.3-Flash ⬇ hf.co/zai-org/GLM-5.3-Flash 27.8 Z AI 2026-08-26
35 MiMo-V2.6-Flash 26.7 Xiaomi 2026-09-21
36 Mistral Large 4 (High) 2 effort-varianter 26.1 Mistral 2026-10-29
37 Mistral Large 4 Preview 25.3 Mistral 2026-10-06
38 Muse Spark Open (Xhigh) 24.9 Meta 2026-10-15
39 Agnes 2.5 Pro Theta 24 Sapiens AI 2026-11-07
40 GPT-5.6 Luna (Max) 6 effort-varianter 23.8 OpenAI 2026-07-09

Hugging Face — trending familier

#Rod-repoTrendingLikes DownloadsParamsAlder
1 Jackrong/DeepSeek-V4-Pro-Qwen3.5-9B +1 derivater 3756 6345 21586 9.4B 15 d
2 NousResearch/Arete-27B 419 559 5545 27.4B 7 d
3 Lightricks/LTX-2.5 411 11804 2778909 114 d
4 Qwen/Qwen3.8-27B +11 derivater 405 19900 6960395 27.8B 101 d
5 speridlabs/idol 382 2304 31367 16 d
6 moonshotai/Kimi-K3.1-Pro 334 2040 40953 2780.1B 12 d
7 Zyphra/ZONOS3-Mini 261 301 13068 5 d
8 thinkingmachines/Inkling-2-Small 240 253 1064 594.1B 9 d
9 Qwen/Qwen3.5-4B-Base +2 derivater 239 1099 392582 4.5B 20 d
10 google/gemma-4-31B +1 derivater 214 264 105256 7 d
11 facebook/sam3 204 6541 2092498 859.9M 372 d
12 Cactus-Compute/mirage 203 261 18959 8 d
13 ConwayResearch/Underdog-Woof 199 247 10 8 d
14 ricursiveAI/Ripple-65B-Preview 196 572 8500 64.8B 10 d
15 deepseek-ai/DeepSeek-V4.1-Flash-1017 +2 derivater 180 5655 398161 763.2B 28 d

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