MiniCPM-V 4.7 35B-A3B has 34.66 billion total parameters. The selected Q4_K_M GGUF file is 21.166 GB.
Data snapshot: . 3 published tests in this catalog.
Model file and details
Provider
OpenBMB
Total parameters
34,659,889,792
Exact file size
21,166,255,008 bytes
Quantization / precision
Q4_K_M
Weight repository
openbmb/MiniCPM-V-4.7-35B-A3B-gguf
Original model repository
openbmb/MiniCPM-V-4.7-35B-A3B
Weight revision
3c3f7e84f14a760ff26d17ab5932d4aa996227c0
Model revision
7dfb75f94371a4738b4ba5994800ed29c74cc812
Runtime notes
Confirm model architecture and quantization support in your runner.
Recorded release date
2026-10-12
Date evidence
OpenBMB completed the Apache-2.0 MiniCPM-V 4.7 35B-A3B release card and official GGUF set on 12 October 2026. On 13 October it republished the selected 21,166,255,008-byte Q4_K_M language file and added an optional 789,442,368-byte BF16 DSpark draft GGUF. On 14 October OpenBMB added its recommended instruct-mode generation settings: temperature 0.7, top_p 0.95, top_k 100, min_p 0.0, presence penalty 0 and repetition penalty 1.03. These are publisher usage defaults, not benchmark conditions or evidence of Q4_K_M speed or quality; multimodal use still requires one separate F16 vision projector.
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.
MMBench · reported
91.72 / 100
Publisher-reported MMBench score. The card does not identify the benchmark version or detailed evaluation settings, so this remains separate from versioned results.
Benchmark efficiency: 4.333 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: OpenBMB's comparison chart reports 91.72 for the official MiniCPM-V 4.7 35B-A3B source checkpoint in Instruct mode with 4x visual-token compression.
Publisher-reported source-checkpoint result; not independent and not a validation of the selected Q4_K_M GGUF. The card says results combine public reports and OpenBMB local evaluation but does not identify the MMBench version or full settings. The selected GGUF exposes 34,659,889,792 language parameters and requires a separate projector for vision.
Publisher-reported OCRBench score for multimodal OCR ability. The card does not provide enough protocol detail to compare it as an independent artifact test.
Benchmark efficiency: 4.309 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: OpenBMB's comparison chart reports 91.20 on OCRBench for the official source checkpoint in Instruct mode with 4x visual-token compression.
Publisher-reported source-checkpoint result; not independent and not a validation of the selected Q4_K_M GGUF. The exact OCRBench protocol is not stated. Vision use requires the separate 910,491,648-byte 4x F16 projector in addition to the 21,166,255,008-byte language file, plus runtime and context memory.
Publisher-reported ScreenSpot score for screen grounding. The card does not identify a version or full evaluation protocol.
Benchmark efficiency: 4.11 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: OpenBMB's comparison chart reports 86.99 on ScreenSpot for the official source checkpoint in Instruct mode with 4x visual-token compression.
Publisher-reported source-checkpoint result; not independent and not a validation of the selected Q4_K_M GGUF. The ScreenSpot version and detailed settings are not stated. Current llama.cpp conversion and inference must use the MiniCPM-V 4.7 GGUF contract introduced at or after PR #29416.
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 21.166 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 21.17 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: 2026-10-02. Apple US launch price for the base M5 Max Mac Studio, published 22 September 2026: USD 2,499. 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 21.166 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.