Devstral Small 2 vs ZAYA1-8B

Devstral Small 2 90

Memory 25, the smallest build at 4 bits from a runtime's own library is Ollama's devstral-small-2 at Q4_K_M, 15 GB, and 15 GB × 1.2 = 18 GB, within 24 GB. Mistral's own FP8 files are 25.79 GB. Runtimes 25, Released vLLM 0.33.0, SGLang 0.5.23 and Transformers 5.22.2 have the mistral3 model, llama.cpp b12242 has the architecture with a GGUF from its own organisation, and Ollama's library has devstral-small-2. Builds 10, one quantised precision from Mistral, the FP8 weights it releases. Terms 15, Apache 2.0 in the card's metadata and text, an OSI licence with nothing added. Download 10, the Hub API reports gated false on 10 November 2026, so the files download with no account and nothing to accept. Hosted 5, Mistral's own API serves Devstral Small 2: the card says the model is available through its API and points there when a local runtime underperforms. 25 + 25 + 10 + 15 + 10 + 5 = 90.

ZAYA1-8B 80

Memory 30, the smallest build from Zyphra is the MXFP4-Experts repository, 5.85 GB, and 5.85 GB × 1.2 = 7.0 GB, within 16 GB. Runtimes 10, Released Transformers 5.22.2 has the zaya model and SGLang 0.5.23 has srt/models/zaya.py. Builds 15, Quantised builds from Zyphra at two precisions, FP8 and MXFP4 for the experts, each in its own repository. Terms 15, Apache 2.0 in the card's metadata, an OSI licence with nothing added. Download 10, the Hub API reports gated false for the model and its builds on 10 November 2026, so the files download with no account and nothing to accept. Hosted 0, Hugging Face maps no inference provider to it, and the card names no API. Zyphra's site lists a cloud product for its open models, which wasn't checked for this model. 30 + 10 + 15 + 15 + 10 + 0 = 80.

Devstral Small 2 60

File safety 25, safetensors only, no .bin, .pt or .pth files and no Python in the repository. Provenance 15, files in Mistral's verified mistralai organisation, which Mistral's site links to per our Mistral Large 4 listing (10). Safety evaluation 0, no safety results on the card. Agent guidance 0, the card has no limitations or safety section, and nothing on prompt injection, instruction priority or what the model refuses. Licence risk 15, Apache 2.0 has no use policy, threshold or change clause. Security programme 5, mistral.ai's security.txt gives a security contact and a submission policy, as read for the Mistral Large 4 listing (5). 25 + 15 + 0 + 0 + 15 + 5 = 60.

ZAYA1-8B 55

File safety 25, safetensors only, no .bin, .pt or .pth files and no Python in the repository. Provenance 15, the files sit in the Zyphra organisation, which the card ties to zyphra.com's blog post for the model (10). Safety evaluation 0, no safety results on the card. Agent guidance 0, the card has no limitations or safety section, and nothing on prompt injection, instruction priority or what the model refuses. Licence risk 15, Apache 2.0 has no use policy, threshold or change clause. Security programme 0, no way to report a flaw on the card, and www.zyphra.com has no security.txt. 25 + 15 + 0 + 0 + 15 + 0 = 55.

Devstral Small 2 65

Chat template 20, chat_template.jinja in the repository (11,362 characters), with the tokenizer's tekken file for mistral-common. Tool-calling format 25, the card documents tools in the OpenAI function format with an example, names the mistral tool-call parser with auto tool choice for vLLM, shows the same for SGLang, and ships test prompts with the expected tool-call output. Trained tool calling 10, the card says the model excels at using tools, and its tests cover one tool call and calls made one at a time in sequence. Structured output 0, Nothing on the card about JSON or schema-constrained answers for this model. Reasoning control 0, no reasoning mode is documented, and a June 2026 commit removed stray thinking tokens from the files. Context 10, 256K tokens per the card, served at 262,144 in its vLLM command. 20 + 25 + 10 + 0 + 0 + 10 = 65.

ZAYA1-8B 60

Chat template 20, chat_template.jinja in the repository (10,700 characters). Tool-calling format 10, only the template and a fork say it. Trained tool calling 10, the template takes tools and renders calls, and the card recommends settings for agent use, but it doesn't say the model was trained for tool calling or for several calls in a turn, so the lower band. Structured output 0, Nothing on the card about JSON or schema-constrained output. Reasoning control 10, the template has an enable_thinking switch, and the card names the qwen3 reasoning parser, which returns the reasoning apart from the answer. Context 10, 131,072 positions in config.json, for an architecture Zyphra trained end to end, so the config is the model's own. 20 + 10 + 10 + 0 + 10 + 10 = 60.