Satella-RAG 0.8B has 0.773 billion total parameters. The selected Q5_K_M GGUF file is 0.593 GB.
Data snapshot: . 1 published test in this catalog.
Model file and details
Provider
EphAsad
Total parameters
772,845,888
Exact file size
592,636,544 bytes
Quantization / precision
Q5_K_M
Weight repository
EphAsad/SatellaRAG-0.8B
Original model repository
EphAsad/SatellaRAG-0.8B
Weight revision
aff60d6dd5c92ff5e0eb6ebdc8022dffca499010
Model revision
aff60d6dd5c92ff5e0eb6ebdc8022dffca499010
Runtime notes
Confirm model architecture and quantization support in your runner.
Recorded release date
2026-10-12
Date evidence
The Apache-2.0 repository and same-publisher GGUF files were created on 12 October 2026. The selected Q5_K_M language file is the card's practical balance recommendation; the optional 207,346,048-byte vision projector is separate and the release evaluation is text-only.
These scores were compiled from the linked sources. These published scores were not independently tested here. Different prompts, precision, thinking budgets and tools can change results.
Satella RAG · overall reference F1
76.24 / 100
Publisher project-specific word-overlap F1 on a 120-example held-out RAG comparison. Examples came from dataset families also used for training, so this is not an independent general-RAG benchmark.
Benchmark efficiency: 128.645 published score points per GB of selected weights. This ratio is not an inference-speed measure; check the test settings to see whether the score covers the selected quantization.
Test settings: Publisher initial held-out comparison on 120 RAG examples, using word-overlap reference F1. Held-out splits were created before training-mixture sampling, exact prompt collisions were removed, and overlength examples were rejected rather than truncated. Satella scored 0.7624 versus 0.3270 for Qwen3.5-0.8B.
Publisher-run project-specific result, not independent evaluation and not a direct test of the selected Q5_K_M file. Evaluation examples came from dataset families also used for training; exact decontamination does not rule out semantic overlap. Reference word F1 is wording-sensitive, the 120-example comparison had no valid Spark subset, and standardized unanswerable examples do not prove general hallucination resistance. The card reports a separate small 10-run diagnostic where thinking mode answered all tested runs correctly and non-thinking mode answered 2/10, but explicitly says it is not a general benchmark. Reasoning should remain enabled. The 592,636,544-byte language file excludes the optional projector, runtime, KV cache and context memory and does not establish hardware compatibility or speed. Apache-2.0 license.
Context and version evidence: The repository and file name identify the selected model. Revision hashes are shown above when recorded. Tested context length, full hardware configuration and run date are not consistently recorded in this snapshot. Only settings explicitly stated in each benchmark record are known.
The 0.593 GB figure covers model weights in this file. It excludes the context cache, runtime, operating system and any separate multimodal files. It is not a measured RAM requirement or a token-generation speed.
These complete configurations pass the initial weight-size check for the selected 0.59 GB file. Allow extra memory for context and software. Check the model’s runtime notes and your inference software before buying.
Listing evidence: 2026-09-25. Amazon.com base price visible in an owner-supplied screenshot reviewed 25 Sep 2026; capture time not supplied. Dated reference, not a live quote. Excludes card signup offers, coupons, tax and shipping.
Listing evidence: 2026-10-03. Apple US launch price announced 22 September 2026: USD 1,699. Dated manufacturer reference, not an Amazon offer; tax and shipping may differ.
Listing evidence: 2026-09-24. GMKtec, EVO-X2 / 64GB / 1TB, recorded 22 Sep 2026; exact ASIN capacity identified 24 Sep. This is a dated reference, not a current Amazon price; tax, shipping and bundle contents may differ.
Memory matches are estimates, not tested compatibility or speed. As an Amazon Associate I earn from qualifying purchases.
Confirm model architecture and quantization support in your runner. The selected weights use 0.593 GB; this excludes context cache, runtime buffers and other programs.
These are the first listed tiers that pass our reserve policy, not measured minimum requirements. System RAM and VRAM remain separate. Use the memory tool to change the reserve for your context and runner.