News-Tracker

Modeller

AI & LLM · daglig briefing

Ændringer siden forrige snapshot

Ny i indekset

  • Nemotron 3.5 Lightning er ny i Artificial Analysis' indeks (intelligence 23.6, NVIDIA).
  • Nova 2.0 Lite (high) er ny i Artificial Analysis' indeks (intelligence 20.8, Amazon).

Score ændret

  • Muse Spark: intelligence 43.5 → 44.3 (+0.8).
  • KAT Coder Pro V2: intelligence 33.9 → 34.5 (+0.6).
  • DeepSeek V3.1 Terminus (Reasoning): intelligence 30.6 → 31.1 (+0.5).
  • Phi-4 Mini Instruct: intelligence 6.2 → 5.7 (-0.5).
  • MiniCPM-V 4.6 1.3B: intelligence 4.3 → 3.8 (-0.5).

Ny i HF-trending

  • Lightricks/LTX-2.5 er ny i Hugging Faces trending (score 258).
  • nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-NVFP4 er ny i Hugging Faces trending (score 140, 1 derivater).
  • Cactus-Compute/needle2 er ny i Hugging Faces trending (score 55).
  • ReadyArt/gemma-4-31B-it-scotoma-2 er ny i Hugging Faces trending (score 42).
  • huihui-ai/Huihui-CyberStrike-OffSec-35B-abliterated er ny i Hugging Faces trending (score 42).

Intelligence Index — Artificial Analysis

#ModelIntel.Agentic CreatorUdgivet
1 Claude Opus 5 (Adaptive Reasoning, Max Effort) 5 effort-varianter 63.1 59.2 Anthropic 2026-07-24
2 Claude Fable 5 (Adaptive Reasoning, Max Effort, Opus 4.8 Fallback) 62.1 56.6 Anthropic 2026-06-09
3 GPT-5.6 Sol (max) 6 effort-varianter 60.9 57.8 OpenAI 2026-07-09
4 Kimi K3 (max) 2 effort-varianter ⬇ hf.co/moonshotai/Kimi-K3 59.7 54.3 Kimi 2026-07-16
5 Qwen3.8 Max 58.1 58.4 Alibaba 2026-08-03
6 Claude Opus 4.8 (Adaptive Reasoning, Max Effort) 57.3 49.4 Anthropic 2026-05-28
7 Muse Spark 1.2 (xhigh) 56.8 49.3 Meta 2026-08-05
8 GPT-5.6 Terra (max) 6 effort-varianter 56.6 50.2 OpenAI 2026-07-09
9 GPT-5.5 (xhigh) 5 effort-varianter 56.3 47.4 OpenAI 2026-04-23
10 Grok 4.5 (high) 55.8 48.9 SpaceXAI 2026-07-08
11 Claude Sonnet 5 (Adaptive Reasoning, Max Effort) 6 effort-varianter 55.3 49.7 Anthropic 2026-06-30
12 Claude Opus 4.7 (Adaptive Reasoning, Max Effort) 2 effort-varianter 55 46.3 Anthropic 2026-04-16
13 Muse Spark 1.1 (xhigh) 53.2 39.7 Meta 2026-07-09
14 GPT-5.4 (xhigh) 3 effort-varianter 53.1 44.2 OpenAI 2026-03-05
15 GLM-5.2 (max) 2 effort-varianter ⬇ hf.co/zai-org/GLM-5.2 52.6 45.7 Z AI 2026-06-16
16 GPT-5.6 Luna (max) 6 effort-varianter 52.3 46.9 OpenAI 2026-07-09
17 Gemini 3.5 Flash (high) 3 effort-varianter 52 39.7 Google 2026-05-19
18 DeepSeek V4 Flash 0731 (Reasoning, Max Effort) ⬇ hf.co/deepseek-ai/DeepSeek-V4-Flash-0731 51.8 48.4 DeepSeek 2026-07-31
19 Gemini 3.6 Flash (high) 51.6 40.5 Google 2026-07-21
20 Claude Sonnet 4.6 (Adaptive Reasoning, Max Effort) 3 effort-varianter 48.4 42.1 Anthropic 2026-02-17
21 Gemini 3.1 Pro Preview 47.7 23 Google 2026-02-19
22 Qwen3.7 Max 46.7 30.9 Alibaba 2026-05-19
23 GPT-5.3 Codex (xhigh) 45.5 OpenAI 2026-02-05
24 MiniMax-M3 ⬇ hf.co/MiniMaxAI/MiniMax-M3 45.4 36.1 MiniMax 2026-06-01
25 DeepSeek V4 Pro (Reasoning, Max Effort) 3 effort-varianter ⬇ hf.co/deepseek-ai/DeepSeek-V4-Pro 45.3 37.8 DeepSeek 2026-04-24
26 Motif 3 (Beta) ⬇ hf.co/Motif-Technologies/Motif-3 45.3 Motif Technologies 2026-07-14
27 Kimi K2.6 2 effort-varianter 45.1 31.2 Kimi 2026-04-20
28 Claude Opus 4.6 (Adaptive Reasoning, Max Effort) 2 effort-varianter 44.9 Anthropic 2026-02-05
29 Muse Spark 44.3 Meta 2026-04-08
30 GPT-5.2 (xhigh) 3 effort-varianter 43.3 OpenAI 2025-12-11
31 Kimi K2.7 Code 43 30.3 Kimi 2026-06-12
32 MiMo-V2.5-Pro 2 effort-varianter ⬇ hf.co/XiaomiMiMo/MiMo-V2.5-Pro 42.9 29.5 Xiaomi 2026-04-22
33 Inkling (xhigh) ⬇ hf.co/thinkingmachines/Inkling 42.3 34.1 Thinking Machines 2026-07-15
34 Hy3 ⬇ hf.co/tencent/Hy3 42.2 31.4 Tencent 2026-07-06
35 DeepSeek V4 Flash (Reasoning, Max Effort) 3 effort-varianter ⬇ hf.co/deepseek-ai/DeepSeek-V4-Flash 42.1 33.7 DeepSeek 2026-04-24
36 Nex-N2-Pro ⬇ hf.co/nex-agi/Nex-N2-Pro 42.1 Nex AGI 2026-06-02
37 Claude Opus 4.5 (Reasoning) 2 effort-varianter 41.9 Anthropic 2025-11-24
38 MiMo-V2-Pro 41.4 Xiaomi 2026-03-18
39 GPT-5.2 Codex (xhigh) 41.2 OpenAI 2025-12-11
40 Inkling Small ⬇ hf.co/thinkingmachines/Inkling-Small 41.2 31.9 Thinking Machines 2026-07-30

Hugging Face — trending familier

#Rod-repoTrendingLikes DownloadsParamsAlder
1 MiniMaxAI/MiniMax-H3 +18 derivater 1335 3593 59368 33.1B 15 d
2 meta-models/Muse-Glimmer-30B +3 derivater 1069 1113 0 29.8B 2 d
3 deepseek-ai/DeepSeek-V4-Flash-0731 +2 derivater 662 3162 1048685 304.2B 12 d
4 moonshotai/Kimi-K3 388 10532 1565484 2779.9B 60 d
5 LiquidAI/LFM2.5-2.6B-Base +2 derivater 369 555 93668 2.7B 14 d
6 deepgrove/maple-preview +1 derivater 291 337 2049 20.2B 7 d
7 Lightricks/LTX-2.5 258 270 39 20 d
8 ethanfel/Qwen3-VL-32B-Ultra-Heretic-H3-ComfyUI-INT8-ConvRot 258 466 0 9 d
9 nvidia/NVIDIA-NemotronLabs-VoiceChat-11B 225 331 653 11.1B 13 d
10 inclusionAI/Ling-3.0-flash +1 derivater 197 308 6148 127.5B 9 d
11 endless-frontier/BigBang-v1 165 171 708 36.0B 9 d
12 inclusionAI/Ling-3.0-tiny 161 161 0 7.9B 2 d
13 mistralai/Shieldstral-1.0-3B 149 232 6769 3.8B 27 d
14 nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-NVFP4 +1 derivater 140 142 19250 17.8B 7 d
15 baidu/Unlimited-OCR 115 4023 2892191 3.3B 54 d

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