OpenRouter model
DeepSeek: DeepSeek V3
DeepSeek-V3 is the latest model from the DeepSeek team, building upon the instruction following and coding abilities of the previous versions. Pre-trained on nearly 15 trillion tokens, the reported evaluations...
- Provider
- OpenRouter
- Availability
- Available
- Model ID
- deepseek/deepseek-chat
- Snapshot
- 2026-10-07
DeepSeek: DeepSeek V3 Benchmarks & Agentic Accuracy
Verified primary research & developer preference metrics
What This Means in Real Software Development
- Highly capable 671B parameter Mixture-of-Experts architecture with 37B active parameters per token.
- Achieves 88.5% on MMLU and 89.3% on MATH-500 at near-zero inference cost ($0.14/M input, $0.28/M output).
- Solid 42.0% on SWE-bench Verified, outperforming standard non-reasoning models like GPT-4o.
- Fully open weights available for private hosting on multi-GPU nodes.
Swarm Multi-Agent Topology
Best Paired With: DeepSeek R1 or Claude 3.7 Sonnet
Cost-effective workhorse for multi-agent workflows where hundreds of prompt turns would be cost-prohibitive on proprietary APIs.
Full compatibility with Swarms multi-agent DAG runner over OpenRouter or self-hosted vLLM/SGLang endpoints.
What Developers & Practitioners Are Saying
Real-world field observations, production strengths, and developer consensus
"DeepSeek V3 (Chat) is the undisputed king of open-weights coding value. At $0.27/M input, it matches or beats GPT-4o across almost every software engineering benchmark."
Verified Real-World Strengths
- Incredible economics ($0.27/M input, $1.10/M output, $0.07/M cached read)
- Solid 42.0% on SWE-bench Verified without thinking overhead
- Superb multilingual programming proficiency (Python, Go, Java, Rust)
Production Caveats & Gaps
- Public API endpoints can experience rate-limiting spikes during peak hours
- Formatting can occasionally drift on very long conversations
Primary Sources & Methodology Citations
- DeepSeek-V3 Technical Report — DeepSeek AI (arXiv:2412.19437) (December 2024)
DeepSeek: DeepSeek V3 model overview
DeepSeek: DeepSeek V3 is listed in the public Swarm catalog through OpenRouter. This page summarizes the verified catalog fields used by Swarm without claiming real-time access for every provider account.
- Provider
- OpenRouter
- Family
- OpenRouter
- Context window
- 163.8K tokens
- Thinking support
- Not supported
- Availability status
- Available
- Catalog ID
- openrouter/deepseek/deepseek-chat
Run DeepSeek: DeepSeek V3 in Swarm
Verified CLI workflow & multi-agent architecture
swarm ctl model set --provider openrouter --model openrouter/deepseek/deepseek-chat# Or switch interactively inside the terminal TUI:
/modelsSubagent Role Matrix
Assign DeepSeek: DeepSeek V3 to specialized system roles in ~/.swarm/config.json:
- Primary / Planner: Orchestrates full task DAGs and multi-stage pipelines.
- Coder Subagent: Generates code and unit tests inside private Git worktrees.
- Finder Subagent: Performs read-only code search, symbol discovery, and mapping.
Public catalog snapshot. Catalog availability is not a real-time provider status or an access guarantee.