Kimi-Linear-48B-A3B-Instruct
Updated 3 h ago · first seen 11 Sept 2026
model_01M294ZGHZKJ43ETJRE5GJZY1Z
- Parameters
- 49.1B · 3B active
- T2 · 5 h ago
- Context
- 1M tokens
- T2 · 4 h ago
- Released
- 30 Oct 2025
- T2 · 4 h ago
- License
- mit
- T2 · 4 h ago
Specification
- Release date
- 30 Oct 2025
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 4 h agomedium
- Openness
- open-weights
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 5 h agomedium
- License
- mit
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 4 h agomedium
- Architecture
- KimiLinearForCausalLM
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 4 h agomedium
- Parameters
- 49.1B
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 5 h agomedium
- Active parameters
- 3B
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 4 h agomedium
- Context window
- 1M tokens
Source:Artificial AnalysisT2observed 4 h agomedium
- File size
- 98.3 GB
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 4 h agomedium
- Hugging Face repo
- moonshotai/Kimi-Linear-48B-A3B-Instruct
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 5 h agomedium
- Pipeline tag
- text-generation
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 5 h agomedium
- Model card
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 5 h agomedium
- Downloads
- 176,764
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 3 h agomedium
- Likes
- 595
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 3 h agomedium
- Gated
- No
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 4 h agomedium
- Hf inference providers
- featherless-ai
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 4 h agomedium
- Last modified
- 2025-12-16T14:50:08+00:00
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 5 h agomedium
- Library name
- transformers
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 4 h agomedium
- Downloads all time
- 1,187,818
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 3 h agomedium
- Model type
- kimi_linear
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 4 h agomedium
- Tags
- transformers, safetensors, kimi_linear, text-generation, custom_code
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 4 h agomedium
- Weights dtype
- BF16, F32
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 4 h agomedium
Each value shows its source, tier and observation time. Conflicting claims are kept side by side and flagged — never averaged. How AI Atlas records facts →
Provenance
Attributed facts
22
Source tiers
T222
Freshest observation
3 h ago
Conflicts
None
Modalities
Modalities unavailable.
Capabilities
Tool calling
Unavailable
Structured output
Unavailable
Reasoning
No
Artificial Analysis · T2
Vision
Unavailable
Audio
Unavailable
Fine-tuning available
Unavailable
- Context window
- 1M tokens
Source:Artificial AnalysisT2observed 4 h agomedium
| Benchmark | Score | Metric | Config | Evaluated | Source |
|---|---|---|---|---|---|
| τ²-bench | 0% | pass^1 | aa_slug=kimi-linear-48b-a3b-instruct · evaluator=Artificial Analysis · index_version=4.3 | — | artificialanalysis.aiT2 |
| Artificial Analysis Intelligence Index | 7.3 | index | aa_slug=kimi-linear-48b-a3b-instruct · version=4.3 · estimated=true | — | artificialanalysis.aiT2 |
| GPQA | 41.21% | accuracy | aa_slug=kimi-linear-48b-a3b-instruct · variant=GPQA Diamond · evaluator=Artificial Analysis · index_version=4.3 | — | artificialanalysis.aiT2 |
| Humanity's Last Exam | 2.46% | accuracy | aa_slug=kimi-linear-48b-a3b-instruct · evaluator=Artificial Analysis · index_version=4.3 | — | artificialanalysis.aiT2 |
| IFBench | 28.1% | accuracy | aa_slug=kimi-linear-48b-a3b-instruct · evaluator=Artificial Analysis · index_version=4.3 | — | artificialanalysis.aiT2 |
| Terminal-Bench | 11.36% | accuracy | aa_slug=kimi-linear-48b-a3b-instruct · variant=hard · evaluator=Artificial Analysis · index_version=4.3 | — | artificialanalysis.aiT2 |
Scores are reported as published, with their evaluation configuration (harness, prompting, judge). Results with different configs are not directly comparable — see methodology.
Current prices
No current prices recorded
Price history
Memory need = bytes per parameter (4-bit ≈ 0.5 × 1.15 overhead, 8-bit 1.0, fp16 2.0) + a KV-cache allowance. Not a measurement.
| Hardware | Quantization | Memory | Est. need | Fits |
|---|---|---|---|---|
| NVIDIA DGX B200 | 4bit | 1,440 GB | 28.8 GB | Yes |
| Apple M3 Ultra | 4bit | — | 28.8 GB | Yes |
| Mac Studio (Apple M5 Ultra) | 4bit | — | 28.8 GB | Yes |
| AMD Instinct MI325X | 4bit | 256 GB | 28.8 GB | Yes |
| AMD Instinct MI300X | 4bit | 192 GB | 28.8 GB | Yes |
| Apple M2 Ultra | 4bit | — | 28.8 GB | Yes |
| NVIDIA B200 | 4bit | 180 GB | 28.8 GB | Yes |
| NVIDIA H200 | 4bit | 141 GB | 28.8 GB | Yes |
| NVIDIA H200 NVL | 4bit | 141 GB | 28.8 GB | Yes |
| Apple M1 Ultra | 4bit | — | 28.8 GB | Yes |
| Apple M3 Max | 4bit | — | 28.8 GB | Yes |
| Apple M4 Max | 4bit | — | 28.8 GB | Yes |
| Mac Studio (Apple M5 Max) | 4bit | — | 28.8 GB | Yes |
| MacBook Pro (Apple M5 Max) | 4bit | — | 28.8 GB | Yes |
| NVIDIA DGX Spark | 4bit | 128 GB | 28.8 GB | Yes |
| Apple M2 Max | 4bit | — | 28.8 GB | Yes |
| NVIDIA H100 NVL | 4bit | 94 GB | 28.8 GB | Yes |
| NVIDIA A100 80GB | 4bit | 80 GB | 28.8 GB | Yes |
| NVIDIA H100 SXM | 4bit | 80 GB | 28.8 GB | Yes |
| Apple M1 Max | 4bit | — | 28.8 GB | Yes |
| Apple M4 Pro | 4bit | — | 28.8 GB | Yes |
| Mac mini (Apple M5 Pro) | 4bit | — | 28.8 GB | Yes |
| MacBook Pro (Apple M5 Pro) | 4bit | — | 28.8 GB | Yes |
| Apple M3 Pro | 4bit | — | 28.8 GB | Yes |
| Apple M1 Pro | 4bit | — | 28.8 GB | Yes |
| Apple M2 Pro | 4bit | — | 28.8 GB | Yes |
| Apple M4 | 4bit | — | 28.8 GB | Yes |
| iMac (Apple M4) | 4bit | — | 28.8 GB | Yes |
| Mac mini (Apple M6) | 4bit | — | 28.8 GB | Yes |
| MacBook Air (Apple M5) | 4bit | — | 28.8 GB | Yes |
| MacBook Pro (Apple M5) | 4bit | — | 28.8 GB | Yes |
| NVIDIA GeForce RTX 5090 | 4bit | 32 GB | 28.8 GB | Yes |
| Apple M2 | 4bit | — | 28.8 GB | No |
| Apple M3 | 4bit | — | 28.8 GB | No |
| NVIDIA GeForce RTX 3090 | 4bit | 24 GB | 28.8 GB | No |
| NVIDIA GeForce RTX 4090 | 4bit | 24 GB | 28.8 GB | No |
| Apple M1 | 4bit | — | 28.8 GB | No |
Ancestors 0
None recorded.
This model
Kimi-Linear-48B-A3B-Instruct
49.1B params
Quantizations 0
None recorded.
Descendants 1
Papers 2
- arXiv:2510.26692Active35
- arXiv:2412.06464Active35
Repositories 0
No repositories linked yet.
As of
Rewind the record: see this entity's attributes exactly as AI Atlas knew them on a given day.
Claim history · Weights dtype
Weights dtypeweights_dtype1
| Value | Valid from → to | Status | Source | Confidence | Extractor |
|---|---|---|---|---|---|
| BF16, F32 | → current | current | Hugging Face Hub (public pages, model cards, papers)T2 | medium | deterministic |
Claims are temporal and append-only: a new observation closes the previous claim (valid_to) instead of overwriting it. Conflicting claims from different sources are kept side by side and flagged — never averaged. Methodology →
Kimi-Linear-48B-A3B-Instruct scores 11.36% on Terminal-Bench
artificial_analysisKimi-Linear-48B-A3B-Instruct scores 0% on τ²-bench
artificial_analysisKimi-Linear-48B-A3B-Instruct scores 28.1% on IFBench
artificial_analysisKimi-Linear-48B-A3B-Instruct scores 2.46% on Humanity's Last Exam
artificial_analysisKimi-Linear-48B-A3B-Instruct scores 41.21% on GPQA
artificial_analysisKimi-Linear-48B-A3B-Instruct scores 7.3 on Artificial Analysis Intelligence Index
artificial_analysisNew model: Kimi-Linear-48B-A3B-Instruct (Moonshot AI)
huggingface
| Source | Document | Type | Tier | Last observed | Snapshots |
|---|---|---|---|---|---|
| Hugging Face Hub (public pages, model cards, papers) | huggingface.co/moonshotai/Kimi-Linear-48B-A3B-Instruct/raw/main/README.md | model_card | T2· Quality secondary | 2 h ago | 1 |
| Hugging Face Hub (public pages, model cards, papers) | huggingface.co/moonshotai/Kimi-Linear-48B-A3B-Instruct | model_page | T2· Quality secondary | 3 h ago | 3 |
| Hugging Face Hub (public pages, model cards, papers) | huggingface.co/cerebras/Kimi-Linear-REAP-35B-A3B-Instruct | model_page | T2· Quality secondary | 3 h ago | 3 |
| Hugging Face Hub (public pages, model cards, papers) | huggingface.co/models?author=moonshotai&p=0&sort=downloads | listing | T2· Quality secondary | 3 h ago | 3 |
| Artificial Analysis | artificialanalysis.ai/leaderboards/models | leaderboard | T2· Quality secondary | 4 h ago | 1 |
Tier 1 = official/primary, 2 = quality secondary, 3 = community, 4 = unverified. Every snapshot is archived; see all sources and the methodology.