DeepSeek V4 Flash
DeepSeek V4 Flash is an efficiency-optimized Mixture-of-Experts model with 284B total parameters (13B activated) and a 1M-token context window, designed for fast inference and high-throughput workloads while maintaining strong reasoning and coding performance. Hybrid attention enables efficient long-context processing, with high and xhigh reasoning efforts, making it ideal for coding assistants, chat systems, and agent workflows.
Quick start
Drop-in requests for the OpenAI-compatible Deva endpoint.
1curl https://api.deva.me/v1/chat/completions \2 -H "Authorization: Bearer $DEVA_API_KEY" \3 -H "Content-Type: application/json" \4 -d '{5 "model": "deepseek/deepseek-v4-flash",6 "messages": [{"role":"user","content":"Hello from Deva"}],7 "stream": true8 }'Capabilities
Feature metadata advertised for this model.
Related models
More options from deepseek and the recommended set.
DEEPSEEK: DeepSeek Chat v3.1
DeepSeek-V3.1 is a large hybrid reasoning model (671B parameters, 37B active) supporting both thinking and non-thinking modes. It extends DeepSeek-V3 with two-phase long-context training up to 128K tokens and FP8 inference, improving tool use, code generation, and reasoning efficiency to a level comparable with DeepSeek-R1 on hard benchmarks while responding faster. It supports structured tool calling and code and search agents for research, coding, and agentic workflows.
DEEPSEEK: DeepSeek V4 Pro
DeepSeek V4 Pro is a large-scale Mixture-of-Experts model with 1.6T total parameters (49B activated) and a 1M-token context window, built for advanced reasoning, coding, and long-horizon agent workflows. It uses a hybrid attention system for efficient long-context processing and supports high and xhigh reasoning efforts, suiting full-codebase analysis, multi-step automation, and large-scale synthesis.
X AI: Grok 4.3
Grok 4.3 is a reasoning model from xAI that accepts text and image inputs with text output, suited to agentic workflows, instruction-following, and applications requiring high factual accuracy. Reasoning effort is configurable (none, low, medium, or high), and a 1M-token context window with no output limit makes it well-suited to long-document analysis, deep research, and multi-step agentic tasks.
ANTHROPIC: Claude Opus 4.7
Claude Opus 4.7 is the next generation of Anthropic's Opus family, built for long-running, asynchronous agents. Building on Opus 4.6's coding and agentic strengths, it delivers stronger performance on complex, multi-step tasks such as large codebases, multi-stage debugging, and end-to-end project orchestration, plus improved knowledge work from document drafting to data analysis, maintaining coherence across very long outputs and extended sessions.