Nimbus-9B-v2.1 GGUF

Text-only llama.cpp artifacts for Nimbus-9B-v2.1. Q5_K_M is the quality-first local default; Q4_K_M is listed only where it actually exists.

Choose a Nimbus model

Model	Best fit	Transformers	GGUF
Nimbus-2B	Fast drafting and focused edits	`Nimbus-Labs/Nimbus-2B`	`Nimbus-Labs/Nimbus-2B-GGUF`
Nimbus-4B	Balanced implementation and debugging	`Nimbus-Labs/Nimbus-4B`	`Nimbus-Labs/Nimbus-4B-GGUF`
Nimbus-9B v2.1	Deeper coding and reasoning	`Nimbus-Labs/Nimbus-9B-v2.1`	`Nimbus-Labs/Nimbus-9B-v2.1-GGUF`

The adjacent assets/nimbus-family-footprint.json contains the plotted values. Download size is not runtime memory: context cache and runtime buffers require additional capacity.

Downloads

File	Role	Bytes	SHA-256
`Nimbus-9B-v2.1-BF16.gguf`	BF16 reference	17,920,696,672 (17.92 GB)	`47ef4cd04c506292785131fff06d5d5d502c082347a44380c81cfe3316054166`
`Nimbus-9B-v2.1-Q4_K_M.gguf`	Memory-first	5,629,108,576 (5.63 GB)	`a3db6956cc148f332dd1afdec5ba389efabe02cf73af6cf975866f5557a9d74f`
`Nimbus-9B-v2.1-Q5_K_M.gguf`	Quality-first default	6,467,969,376 (6.47 GB)	`84c8604a77bcccf850e2a89bf2f3a28d2d846bf11e5f8dbec3094fe31be6e4f6`

llama.cpp

Validated release runtime: llama.cpp b10007. Native thinking uses the supplied Qwen/Ornith chat template and DeepSeek-style reasoning parsing.

llama-server --model Nimbus-9B-v2.1-Q5_K_M.gguf --ctx-size 65536 --n-gpu-layers all --reasoning-format deepseek

Evaluation

The released Q5_K_M artifact was evaluated with llama.cpp b10007 on Vulkan, native thinking, one answer per task, temperature 0.6, top-p 0.95, top-k 20, seed 42, and the official network-disabled EvalPlus scorer. The base run allowed 32,768 completion tokens. Each original length-plus-empty nonresponse received exactly one preserved 60,000-token recovery attempt; there was no repeated resampling.

Benchmark	Passed	Total	pass@1
HumanEval	146	164	89.0%
HumanEval+	135	164	82.3%
MBPP	330	378	87.3%
MBPP+	277	378	73.3%

Three HumanEval and seven MBPP truncations recovered at the larger limit. Two HumanEval and three MBPP nonresponses remained zero-valued. “Plus” uses EvalPlus' official base-and-plus pass condition.

The adjacent assets/nimbus-9b-v2.1-evalplus.json is the machine-readable source for this chart.

Release landscape

This is a capability and packaging comparison, not a cross-model score claim. Other publishers' benchmarks use different harnesses, prompts, inference budgets, and hardware, so they are not mixed into the Nimbus EvalPlus chart.

Release	Parameters	Modality	Published context	Positioning
Nimbus-9B v2.1	9B	Multimodal Transformers; text-only GGUF	262,144 validated for this release	Local coding/reasoning; BF16, Q4_K_M, Q5_K_M with checksums
Qwen3.5-9B	9B	Multimodal	262,144 in current serving guidance	Foundation model; reasoning and tool use
Ornith-1.0-9B	9B	Multimodal	262,144 in current serving guidance	Immediate upstream focused on coding agents
Granite-3.3-8B-Instruct	8B	Text	128,000	General instruction, code, function calling, and RAG

Nimbus-9B v2.1 inherits its architecture and foundation capabilities from Qwen3.5 through Ornith. Nimbus' contribution is the post-training, release packaging, quantization, and exact evaluation reported above—not authorship of the foundation model.

Provenance and scope

Quantized from Nimbus-Labs/Nimbus-9B-v2.1
Foundation: Ornith-1.0-9B on Qwen3.5-9B
GGUF files are text-only; the Transformers repository retains the multimodal components.
Do not infer memory fit from file size alone; context KV cache and runtime buffers require additional memory.