Ministral 3 8B

Artificial Analysis ranks Ministral 3 8B among 19 Mistral models evaluated, placing it in the lowest-price tier at $0.15 blended per million tokens, tied with Mistral Small 4 as the fifth-cheapest Mistral offering. The analysis highlights pricing and output-speed characteristics useful for throughput-sensitive and cost In Artificial Analysis' intelligence rankings, Ministral 3 8B falls below the top-scoring models (GLM-5.2 max at 39, Mistral Medium 3.5 at 15, Mistral Small 4 at 11), reflecting its positioning as a compact edge model rather than a frontier-tier system. The page also reports that Ministral 3 3B leads Mistral output spe

The Opper AI Mistral release tracker confirms Ministral 3 8B as a distinct entry in the Mistral lineup, released on December 2, 2025, alongside sibling variants Ministral 3 3B and Ministral 3 14B. The tracker assigns it an intelligence rating of 6, contextualizing it among 31 total Mistral model releases tracked throug The tracker records that 2 of the 31 tracked Mistral models are currently runnable on the Opper platform, situating Ministral 3 8B within the broader Mistral release cadence that also includes Mistral Large 3, Devstral 2 variants, and later Mistral Small 4 and Mistral Medium 3.5. The listing confirms Ministral 3 8B as

Ministral 3 8B is an 8-billion-parameter dense multimodal language model from Mistral AI, released around December 1, 2025 under the Apache 2.0 license. According to the apxml.com specifications page, it supports a 256K token context window, uses an integrated vision encoder alongside its language decoder, and is optim Technically, Ministral 3 8B features a 32-layer transformer with 4,096 hidden dimensions, 32 query attention heads and 8 key-value heads with head dimension 128, an RMSNorm-normalized SwiGLU feed-forward network with intermediate size 14,336, absolute position embeddings with RoPE theta 1,000,000, and a 131,072-token v