{"slug": "osaurus-ships-diplodocus-to-handle-pdfs-and-spreadsheets-on-local-hardware", "title": "Osaurus Ships Diplodocus to Handle PDFs and Spreadsheets on Local Hardware", "summary": "Osaurus released Diplodocus preview 4 as three llama.cpp GGUF builds of a 9B finance and documents assistant fine-tuned from IFM's K2-Horizon-7B under Apache 2.0, with workbook and document task scores rising from 59/51 to 70/70 versus preview 3. The quantizations are Q4_K_M at 6.33 GB, Q6_K at 7.76 GB (recommended) and Q8_0 at 9.57 GB, and loading any of them requires llama.cpp with K2 Horizon support from PR #29535, merged October 6, 2026. Osaurus lists known limits including live formulas sometimes rendering as text and unreliable signed PDF edits, and file operations still need a host application that implements and executes the model's tool calls.", "body_md": "# Osaurus Ships Diplodocus to Handle PDFs and Spreadsheets on Local Hardware\n\nOsaurus released Diplodocus preview 4 as llama.cpp GGUF files, a 9B finance and spreadsheet assistant built on K2-Horizon-7B with tool use.\n\n- Osaurus released [Diplodocus preview 4 GGUF](https://huggingface.co/OsaurusAI/Diplodocus-preview-GGUF?utm_source=alphasignal) , a 9B finance and documents assistant for llama.cpp.\n- Built on IFM's K2-Horizon-7B, Apache 2.0, with native tool use and toggleable reasoning.\n- Three quant sizes: Q4_K_M at 6.33 GB, Q6_K at 7.76 GB (recommended), Q8_0 at 9.57 GB.\n- Requires llama.cpp with K2 Horizon support merged on 2026-10-06; older builds will not load it.\n- Workbook and document task scores moved from 59/51 to 70/70 versus preview 3.\n- Known limits: live formulas sometimes come out as text, signed PDF edits are unreliable.\n\nOsaurus has released [Diplodocus preview 4](https://huggingface.co/OsaurusAI/Diplodocus-preview-GGUF?utm_source=alphasignal) as three GGUF builds for llama.cpp. The 7B model targets local workflows involving PDFs, spreadsheets, document edits, and calculator-backed financial arithmetic.\n\nThe model fine-tunes IFM’s K2-Horizon-7B on Osaurus’s workspace tools, including in-place workbook and document editing, a calculator, and optional reasoning. File operations still require a host application that implements those tools and executes the model’s calls. The release uses the Apache 2.0 license and supports llama.cpp-based runtimes such as Ollama, LM Studio, Jan, and Docker Model Runner once they include the required architecture support.\n\n## Three quantizations, one hard dependency\n\nAll three builds use an importance matrix, which applies calibration data during compression to preserve influential weights.\n\n| Build | File size | Intended use | \n|---|---|---|\n| `Q4_K_M` | 6.33 GB | 8 GB GPUs and memory-constrained systems | \n| `Q6_K` | 7.76 GB | Osaurus’s recommended balance of size and fidelity | \n| `Q8_0` | 9.57 GB | Behavior closest to the original BF16 weights | \n\nActual memory use exceeds the file size because the runtime also allocates memory for the context window, KV cache, and other buffers. Longer contexts can therefore require partial CPU offloading even when the model file fits in VRAM.\n\nLoading any build requires llama.cpp with K2 Horizon support from [PR #29535](https://github.com/ggml-org/llama.cpp/pull/29535?utm_source=alphasignal), merged on October 6, 2026. Older llama.cpp builds will reject the `k2-horizon` architecture, including applications that bundle an outdated engine. Mac users running the Osaurus application can instead use the separate JANG_6M MLX package containing the same weights.\n\n## This story is for Pro members\n\nYou've reached the end of the free preview. Upgrade to AlphaSignal Pro to read the full article - and everything else behind the paywall.", "url": "https://wpnews.pro/news/osaurus-ships-diplodocus-to-handle-pdfs-and-spreadsheets-on-local-hardware", "canonical_source": "https://alphasignal.ai/news/osaurus-ships-diplodocus-to-handle-pdfs-and-spreadsheets-on-local-hardware", "published_at": "2026-11-15 09:31:30+00:00", "updated_at": "2026-11-15 09:53:47.800195+00:00", "lang": "en", "topics": ["large-language-models", "ai-tools", "ai-products", "artificial-intelligence"], "entities": ["Osaurus", "Diplodocus preview 4", "K2-Horizon-7B", "IFM", "llama.cpp", "Hugging Face", "Ollama", "LM Studio"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/osaurus-ships-diplodocus-to-handle-pdfs-and-spreadsheets-on-local-hardware", "markdown": "https://wpnews.pro/news/osaurus-ships-diplodocus-to-handle-pdfs-and-spreadsheets-on-local-hardware.md", "text": "https://wpnews.pro/news/osaurus-ships-diplodocus-to-handle-pdfs-and-spreadsheets-on-local-hardware.txt", "jsonld": "https://wpnews.pro/news/osaurus-ships-diplodocus-to-handle-pdfs-and-spreadsheets-on-local-hardware.jsonld"}}