ChatCodingAI codingReasoningMulti-purpose399B
Chat

Arcee Trinity Large Thinking 400B

Trinity · Apache 2.0

Mixture of Experts: 399B total parameters, 13B active per token.

Arcee AI's largest Mixture-of-Experts reasoning model. 399B total parameters with only 13B active per token (256 experts, 4 selected; 1.56% routing, one of the sparsest MoE architectures in production). 512K native context. Trained on 17 trillion tokens across 2,048 B300 GPUs. Post-trained with extended chain-of-thought reasoning and agentic reinforcement learning. Ranks #2 on PinchBench behind Claude Opus-4.6. Requires a GB300 (288 GB) for full in-VRAM inference; even the M4 Ultra 192 GB cannot fit Q3_K_M. US-built, Apache 2.0 licensed.

Parameters
399B
Architecture
MoE (13B active)
Context
524,288 tokens
Released
2026-04-01
Engines
llama.cpp, vLLM, SGLang
Builder Tools
Claude Code, Codex CLI, Continue, Cursor, LM Studio, Windsurf

Parameters

399B

VRAM

295 GB

Context

512K

Formats

3

GPUs

3

Arcee Trinity Large Thinking 400B (399B) requires 295 GB VRAM at recommended quality (Q5_K_M). On NVIDIA Grace Blackwell Ultra GB300, expect approximately 41 tok/s at Q4_K_M.

Source: OwnRig methodology

VRAM (Recommended)

295 GB

Quantization

Q5_K_M

File Size

285 GB

Max Context

512K tokens

Primary Use

Chat

Memory

VRAM Requirements

QualityQuantizationVRAMFile Size
recommendedQ5_K_M295 GB285 GB
efficientQ4_K_M255 GB244 GB
compressedQ3_K_M195 GB185 GB
Scaling

Context Length Impact

KV cache VRAM at Q5_K_M quality. Longer context = more memory.

ContextKV CacheTotal VRAM
2K410 MB295.4 GBexceeds 24 GB
4K819 MB295.8 GBexceeds 24 GB
8K1.6 GB296.6 GBexceeds 24 GB
16K3.2 GB298.2 GBexceeds 24 GB
32K6.4 GB301.4 GBexceeds 24 GB
64K12.8 GB307.8 GBexceeds 24 GB
128K25.6 GB320.6 GBexceeds 24 GB

Compatible GPUs

3 devices
NVIDIA Grace Blackwell Ultra GB300Q4_K_M41 tok/sExcellent
Apple M4 Max (128GB Unified)Q3_K_M1 tok/sNot viable
Apple M4 Ultra (192GB)Q3_K_M3 tok/sNot viable

Showing 3 of 3 entries

Builder Context

Arcee Trinity Large Thinking 400B is commonly used with Claude Code, Codex CLI, Continue, Cursor, LM Studio, Windsurf. For an AI coding workflow, pair it with an embedding model like nomic-embed-text for local RAG.

FAQ

Frequently Asked Questions

How much VRAM does Arcee Trinity Large Thinking 400B need?
Arcee Trinity Large Thinking 400B requires 295 GB VRAM at recommended quality (Q5_K_M). At lower quality settings, it can fit in as little as 195 GB.
What is the best GPU for Arcee Trinity Large Thinking 400B?
The NVIDIA Grace Blackwell Ultra GB300 delivers the best performance for Arcee Trinity Large Thinking 400B, achieving 41 tok/s at Q4_K_M with an excellent rating.
What quantization should I use for Arcee Trinity Large Thinking 400B?
For the best quality, use Q5_K_M (295 GB VRAM). If your GPU has limited VRAM, Q3_K_M (195 GB) is the most efficient option with acceptable quality.
Is Arcee Trinity Large Thinking 400B good for coding?
Yes. Arcee Trinity Large Thinking 400B is used with Claude Code, Codex CLI, Continue, Cursor, LM Studio, Windsurf for local AI coding. For the best coding experience, pair it with an embedding model for local RAG.
All models

Data confidence: estimated. Source

VRAM requirements are calculated from model parameters and may vary by inference engine, context length, and batch size. Performance estimates are based on community benchmarks and should be verified for your specific configuration.Trinity is a trademark of its respective owner. OwnRig is not affiliated with or endorsed by the model creator.