Aloha! 🌺 Today, we are releasing Ornith-1.0, a self-improving family of open-source models for agentic coding.
plamo-3-8b-fin-instruct
pfnet·Instruct·released 8 Nov 2026
ollama run hf.co/mradermacher/plamo-3-8b-fin-instruct-GGUF:Q4_K_MPLaMo-3-8B-Fin-Instruct is a Japanese language model for the financial domain, developed by Preferred Networks, Inc. It is fine-tuned from a PLaMo-3 base model through continued pre-training on financial corpora followed by supervised fine-tuning and direct preference optimization.
How much VRAM does plamo-3-8b-fin-instruct need?
At Q4_K_M plamo-3-8b-fin-instruct needs about 6.1 GB with an 8K context (6.1 GB at Q4, 18 GB at full precision). Pick your GPU to see what fits and how fast it runs.
Needs: ≈6.2 GB · Q4_K_M · 8K context — Pick a GPU to check the fit.
Advanced: context 8K · KV cache 16-bit
Smallest GPU for each quant and context (KV cache 16-bit)
| Quantisation | 8K | 16K | 32K | 64K | 128K | 256K |
|---|---|---|---|---|---|---|
| Q3_K_M | ||||||
| Q4_K_M | ||||||
| Q5_K_M | ||||||
| Q6_K | ||||||
| Q8_0 | ||||||
| FP16 |
≤16 GB card24–32 GB48–96 GBmore
| Quantisation | Total |
|---|---|
| Q2_KHeavy lossSmallest, noticeably worse quality | ≈4.4 GB |
| Q3_K_MNoticeable lossSmall, some quality loss | ≈5.2 GB |
| Q4_K_MrecommendedSmall lossRecommended default for local use | ≈6.2 GB |
| Q5_K_MNear-losslessNear-lossless for most tasks | ≈7.0 GB |
| Q6_KNear-losslessPractically lossless | ≈7.8 GB |
| Q8_0LosslessLossless in practice, twice the size of Q4 | ≈9.9 GB |
| FP16 / BF16LosslessFull precision, as released | ≈18 GB |
Estimates for 8K tokens of context: weights + KV cache + runtime overhead. Speed assumes memory-bound decoding at your GPU’s bandwidth. Pick a GPU above and the fit column checks every row against it. Weights marked “file” use the real GGUF download size; “est.” rows are computed from the parameter count.
How to pick a quant
Take the biggest quant that fits with room left for context. Q4_K_M is the usual sweet spot. Below Q4 quality drops fast, above Q6 you mostly pay memory for nothing. Big models (70B+, large MoE) tolerate low quants better than small ones.
- LosslessF16, Q8_0
- Near-losslessQ6_K, Q5_K_M
- Small lossQ5_K_S, Q4_K_M, IQ4_NL, IQ4_XS, Q4_K_S
- Noticeable lossQ4_0, Q3_K_L, Q3_K_M, IQ3_M
- Heavy lossIQ3_XXS, Q3_K_S, IQ2_M, Q2_K
Typical quality loss, not measured per model. Tiers come from KL divergence against FP16 on Llama 3 8B: llama.cpp perplexity README: LLaMA 3 8B scoreboard (KLD and ΔPPL vs FP16, Wikitext-2); llama.cpp quantize tool: ΔPPL per quant type @ Llama-3-8B
How much context fits?
At 8K context, Q4_K_M needs ≈6.2 GB(4.9 GB weights + 0.5 GB KV cache + 0.8 GB overhead). Pick a GPU above to see how much context fits.
Weights + overheadKV cache
Can the RTX 4060 run plamo-3-8b-fin-instruct? →
Speed vs quality per quant
Smaller quants run faster and fit in less memory, but lose quality.
Fits in VRAMTight fitCPU offload✕ won't runQuality is typical KL divergence vs FP16, not measured for this model.
Real-world speed of plamo-3-8b-fin-instruct
Tokens per second measured by people running it at home. A GPU shows a median once 3 reports agree on the same setup. Reports far from our estimate wait for a review. Measure yours with llama-bench →
No medians yet. Be one of the first 3 to report a setup.
All reports
Run plamo-3-8b-fin-instruct locally
--quantQ4_K_Mdefault; pick your GPU to tailor itmradermacher/plamo-3-8b-fin-instruct-GGUF
ollama run hf.co/mradermacher/plamo-3-8b-fin-instruct-GGUF:Q4_K_M llama-server -hf mradermacher/plamo-3-8b-fin-instruct-GGUF:Q4_K_MDownloads the GGUF on first run. Installed through llama.app? The same flags work with `llama serve`.
lms get mradermacher/plamo-3-8b-fin-instruct-GGUF@Q4_K_MThen load it from the LM Studio app or with `lms load`.
vllm serve pfnet/plamo-3-8b-fin-instructservices:
vllm:
image: vllm/vllm-openai:latest
command: pfnet/plamo-3-8b-fin-instruct
ports:
- "8000:8000"
ipc: host
volumes:
- ~/.cache/huggingface:/root/.cache/huggingface
deploy:
resources:
reservations:
devices:
- driver: nvidia
count: all
capabilities: [gpu]API on port 8000.
from transformers import pipeline
pipe = pipeline("text-generation", model="pfnet/plamo-3-8b-fin-instruct", dtype="auto", device_map="auto")
messages = [{"role": "user", "content": "Explain self-hosting in one paragraph."}]
out = pipe(messages, max_new_tokens=512)
print(out[0]["generated_text"][-1]["content"])Quantised versions and downloads
| Repository | Format | Quant | Size |
|---|---|---|---|
| mradermacher/plamo-3-8b-fin-instruct-GGUF | GGUF | Q2_K | 3.5 GB |
| mradermacher/plamo-3-8b-fin-instruct-GGUF | GGUF | Q3_K_S | 3.8 GB |
| mradermacher/plamo-3-8b-fin-instruct-GGUF | GGUF | Q3_K_M | 4.3 GB |
| mradermacher/plamo-3-8b-fin-instruct-GGUF | GGUF | Q3_K_L | 4.7 GB |
| mradermacher/plamo-3-8b-fin-instruct-GGUF | GGUF | IQ4_XS | 4.8 GB |
| mradermacher/plamo-3-8b-fin-instruct-GGUF | GGUF | Q4_K_S | 5.0 GB |
| mradermacher/plamo-3-8b-fin-instruct-GGUF | GGUF | Q4_K_M | 5.3 GB |
| mradermacher/plamo-3-8b-fin-instruct-GGUF | GGUF | Q5_K_S | 5.9 GB |
| mradermacher/plamo-3-8b-fin-instruct-GGUF | GGUF | Q5_K_M | 6.1 GB |
| mradermacher/plamo-3-8b-fin-instruct-GGUF | GGUF | Q6_K | 7.0 GB |
About plamo-3-8b-fin-instruct
PLaMo-3-8B-Fin-Instruct is a Japanese language model for the financial domain, developed by Preferred Networks, Inc. It is fine-tuned from a PLaMo-3 base model through continued pre-training on financial corpora followed by supervised fine-tuning and direct preference optimization. This is an instruction-tuned model for the financial domain. The model follows the PLaMo-3 chat format. The model is provided as a chat model that reasons in Japanese: it first writes its thought process in Japanese and then gives the answer. It also supports function calling (tool use). Part of the training was carried out on MN-Core 2, a deep learning accelerator developed by Preferred Networks.
Full model card on Hugging Face →Alternatives to plamo-3-8b-fin-instruct
All alternatives →Chirp Chirp! 🐦 We are introducing Ornith-1.5, a major step toward building foundation models through end-to-end self-improvement.
Qwen3 is the latest generation of large language models in Qwen series, offering a comprehensive suite of dense and mixture-of-experts (MoE) models.
Granite is a family of open-source large language models developed by IBM, designed for enterprise and research use. Granite models are built to be versatile, safe, and efficient — covering a range of sizes and capabilities from compact edge-deployable models to large-scale reasoning systems.
Model Summary: Granite-4.1-8B is a 8B parameter long-context instruct model finetuned from Granite-4.1-8B-Base using a combination of open source instruction datasets with permissive license and internally collected synthetic datasets.
NVIDIA-Nemotron-3.5-Bolt-E4B-BF16 is a small language model (SLM) further post-trained from Gemma-4-E4B-it by NVIDIA. This is the full-precision (BF16) release.
plamo-3-8b-fin-instruct FAQ
- How much VRAM does plamo-3-8b-fin-instruct need?
- About 6.1 GB at Q4_K_M with an 8K context. Q8_0 needs about 9.9 GB and full precision about 18 GB. Longer contexts need more memory for the KV cache.
- Can plamo-3-8b-fin-instruct run on the RTX 4060?
- Yes. It fits at Q5_K_M (about 7.0 GB), at an estimated 25 tokens/s. It fits comfortably on the Apple M4 (16 GB), Radeon RX 6700 XT, RTX 3060 12GB.
- Can I use plamo-3-8b-fin-instruct commercially?
- Yes, with conditions. It is released under plamo-community-license. Custom licence: read the model card before using it commercially.
- What is the context length of plamo-3-8b-fin-instruct?
- 262,144 tokens (256K). The full window needs far more memory than an 8K context.
- How do I run plamo-3-8b-fin-instruct with Ollama or llama.cpp?
- With Ollama: ollama run hf.co/mradermacher/plamo-3-8b-fin-instruct-GGUF:Q4_K_M. With llama.cpp: llama-server -hf mradermacher/plamo-3-8b-fin-instruct-GGUF:Q4_K_M, which starts an OpenAI-compatible server.
Rate this model
Data from Hugging Face, the model card and public leaderboards. Memory and speed figures are estimates; real numbers depend on the runtime, context length and drivers. Always check the licence yourself before commercial use.
0 comments