DeepSeek R1 / V3 family
DeepSeek (Hangzhou-based, Liang Wenfeng founder) — released DeepSeek-V3 671B MoE (37B active) Dec-2024, then R1 reasoning model Jan-2025 that matched o1 at ~3% of inference cost. Caused 'DeepSeek moment' in markets Jan-27-2025 — NVIDIA dropped 17%. MIT license. Substrate for cost-conscious OSS deployers and an active reasoning-research baseline.
Active · Verified 2026-10-31 · fresh as of 2026-10-31 · 58 cited facts, 9 with a third-party source · 1 recorded change · Suggest a correction
At a glance
- Tier
- T1
- Created
- 2023-07 (DeepSeek founded); 2024-05 (V2); 2024-12 (V3); 2025-01 (R1); 2025-08 (V3.1) vendor fix
- Latest release
- DeepSeek-V4.1-Flash-1017 (2026-10-17); DeepSeek-V4-Pro GA (2026-08-13) vendor fix
- License
- MIT (R1 + V3 weights) vendor fix
- GitHub
- github.com/deepseek-ai/DeepSeek-V3 ~94k stars; github.com/deepseek-ai/DeepSeek-R1 ~88k stars vendor fix
- Pricing
- API per 1M in (cache miss)/out, peak: V4-Pro $1.32/$3.96, V4.1-Flash $0.30/$1.20; off-peak 50% lower; weights free vendor fix
- Funding
- Self-funded via High-Flyer (parent quant fund); no external VC disclosed; estimated <$100M training cost (vs $1B+ for peers) vendor fix
- Freshness
- 2026-10-31 fix
- Layer
- large-models repo fix
- Licensing
- open-source repo fix
- Protocols
- openai-compatible vendor fix
When to use
Optimised for: cost-efficient frontier reasoning; OSS-friendly deployment; distillation source for smaller models
Anti-fit: China data-residency for regulated Western workloads; export-control posture; less polished tool-use than Claude/GPT; multimodal less mature
Pros & cons
Pros
MIT-licensed frontier weights; ~3% the inference cost of peers; R1 matches o1 on math/reasoning; spawned a wave of cheaper distilled reasoners.
Cons
China-based — geopolitical and regulatory risk for Western enterprises; tool-use less polished; multimodal lagging; dataset / training-data transparency limited.
Claims & capabilities
R1 matches o1 on AIME/MATH at ~3% the cost; V3 671B MoE / 37B active / 128k context; MIT license (one of most permissive for frontier weights); used for distillation into smaller Qwen / Llama variants
Technical surface
- API surface
- OpenAI-compatible REST + SDKs; HuggingFace Transformers
- Backend storage
- not applicable — substrate foundation model
- Deployment
- Self-hostable (MIT weights) + hosted DeepSeek API + Together / Fireworks / HuggingFace inference
- Embedding model
- not applicable — not a memory product
- Multi-tenancy
- not applicable — substrate foundation model
- MCP
- not applicable — substrate foundation model
- A2A
- not applicable — substrate foundation model
- OpenTelemetry
- not applicable — substrate foundation model
Similar systems
Other foundation models (substrate reference) in the catalog, ranked by inbound references.
- Anthropic Claude (foundation models) T1
Claude family of frontier foundation models — Sonnet 4 (workhorse), Opus 4.5 (frontier capability), Haiku 4.5 (fast/cheap). Frontier model used as substrate by Claude Code, Anthropic Memory tool, Cline, Aider, Continue.dev, Cursor, Goose, and dozens of other catalog entries. Memory in Claude is parametric (in weights) — discrete episodic memory lives outside the model in the Claude Memory tool (see Platform-provider memory section).
- OpenAI GPT family (GPT-5 / GPT-4o / o3 / o4) T1
OpenAI's foundation model family — GPT-5 (2025-08 flagship), GPT-4o (multimodal workhorse), o3/o4 reasoning series. Substrate for ChatGPT, the OpenAI API, Microsoft Copilot, Azure OpenAI, and many downstream agents in this catalog (Devin, Codex CLI, Operator, Cursor mode). Parametric memory in weights; discrete user memory ships in ChatGPT Memory feature (separate row).
- Google Gemini 3 family T1
Google DeepMind's Gemini 3 frontier model family — Gemini 3 Pro (top tier), 3 Flash (fast), 3 Nano (on-device). Substrate for Google Workspace AI, Vertex AI Agent Platform, Gemini Code Assist, Project Mariner, and Project Astra. Multimodal-native; 1M-10M context window. Largest context window of any frontier model. Used by Gemini Enterprise Agent Platform Memory Bank (separate row).
- Mistral Large 2 / Mixtral family T1
French frontier-model lab (Paris). Family includes Mistral Large 2 (123B dense, Jul-2024), Mixtral 8x22B (open-weights MoE), Mistral Small 3 (24B, open-weights Jan-2025), Codestral (code specialist). EU-headquartered alternative to US/China model labs; GDPR-friendly data residency. Distinct from the Mistral Le Chat memory feature (already in catalog) — this row covers the model family itself as substrate.
- xAI Grok 4 T1
Elon Musk's xAI frontier model — Grok 4 launched 2025-07. Trained on Colossus (100k+ GPU supercluster in Memphis). Deeply integrated with X/Twitter (real-time tweet data access). Grok 4 Heavy variant for premium tier. No frontier-tier open-weights — Grok 1 (314B MoE) is the only weight release. The x.ai site now brands itself SpaceXSI (page title, checked 2026-10-29).
- Cohere Command R+ / Command A T1
Toronto-based foundation-model lab (Aidan Gomez, ex-Google Brain 'Attention Is All You Need' co-author). Command family is enterprise / RAG-optimised: Command A (Mar-2025, 111B dense, 256k ctx), Command R+ (104B MoE), Command R7B (smallest tier). Announced merger with Aleph Alpha April-2026. Open weights via CC-BY-NC (non-commercial); Cohere-hosted for commercial.
Sources (57 cited facts, 9 third-party)
Each fact links to the page it came from; the badge says who published it (how to read sources).
- What it is Open-weights reasoning-first frontier model family (R1 / V3 / V3.1) vendor
- What it does & what's distinct DeepSeek (Hangzhou-based, Liang Wenfeng founder) — released DeepSeek-V3 671B MoE (37B ac… vendor
- Claims / benefits R1 matches o1 on AIME/MATH at ~3% the cost; V3 671B MoE / 37B active / 128k context; MIT… vendor
- Modalities text, code; image (DeepSeek-VL line); not yet audio/video at R1/V3 level vendor
- Created 2023-07 (DeepSeek founded); 2024-05 (V2); 2024-12 (V3); 2025-01 (R1); 2025-08 (V3.1) vendor
- Latest release DeepSeek-V4.1-Flash-1017 (2026-10-17); DeepSeek-V4-Pro GA (2026-08-13) vendor
- License MIT (R1 + V3 weights) vendor
- GitHub github.com/deepseek-ai/DeepSeek-V3 ~94k stars; github.com/deepseek-ai/DeepSeek-R1 ~88k s… vendor
- Mindshare Jan-27-2025 'DeepSeek moment' — NVIDIA -17%; LMSYS Arena top-10; massive HuggingFace dow… vendor
- Funding Self-funded via High-Flyer (parent quant fund); no external VC disclosed; estimated <$10… vendor
- Customers / scale OSS deployers via HuggingFace, Together AI, Fireworks; integrated in Perplexity, Continu… vendor
- Pricing API per 1M in (cache miss)/out, peak: V4-Pro $1.32/$3.96, V4.1-Flash $0.30/$1.20; off-pe… vendor
- Compliance China-based; not certified to US frameworks (SOC 2 / HIPAA); export controls + data-resi… vendor
- Data handling Self-hosting recommended for sensitive data (China API may train on submissions); MIT we… vendor
- Deployment Self-hostable (MIT weights) + hosted DeepSeek API + Together / Fireworks / HuggingFace i… vendor
- HQ Hangzhou, China vendor
- Founders / pedigree DeepSeek (Liang Wenfeng; spin-off from quant fund High-Flyer Capital) vendor
- Performance MMLU-Pro: V3 ~76%; AIME 2024: R1 ~80% (matches o1); LiveCodeBench: R1 ~65%; MATH-500: R1… vendor
- Code/weights release MIT — weights public vendor
- API surface OpenAI-compatible REST + SDKs; HuggingFace Transformers vendor
- LLM lock open (MIT license; many hosts) vendor
- Runtimes any (HuggingFace Transformers, vLLM, llama.cpp, MLX) vendor
- Validated verticals OSS deployers, cost-sensitive workloads, academic reasoning research, distillation pipel… vendor
- Time to running <10min via hosted DeepSeek API; <1hr via Together / Fireworks; <1day via on-prem (671B r… vendor
- Anti-fit China data-residency for regulated Western workloads; export-control posture; less polis… vendor
- Optimised for cost-efficient frontier reasoning; OSS-friendly deployment; distillation source for smal… vendor
- Adjacent infrastructure HuggingFace; Together AI; Fireworks; Perplexity / Continue.dev as model option vendor
- Pros MIT-licensed frontier weights; ~3% the inference cost of peers; R1 matches o1 on math/re… vendor
- Cons China-based — geopolitical and regulatory risk for Western enterprises; tool-use less po… vendor
- Project & sources www.deepseek.com/ vendor
- obs-langsmith yes third-party
- obs-datadog yes third-party
- obs-helicone yes third-party
- obs-weave yes third-party
- obs-langfuse yes third-party
- obs-custom DeepSeek API is OpenAI-compatible; picks up OpenAI-proxy observability tools vendor
- cost-token-budget yes vendor
- cost-prompt-caching yes vendor
- cost-semantic-caching no vendor
- cost-batching no vendor
- cost-model-routing no vendor
- cost-streaming-only yes vendor
- cost-observability-cost-attribution no vendor
- eval-langsmith-evals yes third-party
- eval-braintrust yes third-party
- eval-weights-and-biases-agent yes third-party
- eval-helicone-evals yes third-party
- eval-custom-test-harness no vendor
- eval-human-loop no vendor
- eval-production-traffic-replay no vendor
- capability-benchmark-sources swe-bench-verified: 80.6% (DeepSeek-V4-Pro-Max, 2026-04-26) https://arxiv.org/abs/2606.1… paper
- architecture-deployment-locus hybrid repo
- licensing-bucket open-source repo
- positioning-platform-vs-product both repo
- taxonomy-three-layer-rollup large-models repo
- sdk-languages rest vendor
- protocol-support openai-compatible vendor
What changed
Changes to this record's lifecycle, licence, release, pricing and hosting facts, from the catalog's git history. Subscribe (Atom)
- 2026-10-26
- Latest release “V3.1 (2025-08), R1 (2025-01)” → “DeepSeek-V4.1-Flash-1017 (2026-10-17); DeepSeek-V4-Pro GA (2026-08-13)” source
- Pricing “API: V3 $0.27/$1.10 per 1M tokens; R1 $0.55/$2.19; ~10-30x cheaper than Anthropic/OpenAI; weights free under MIT” → “API per 1M in (cache miss)/out, peak: V4-Pro $1.32/$3.96, V4.1-Flash $0.30/$1.20; off-peak 50% lower; weights free” source
Related systems
Referenced by (27)
- Agent Router (formerly Envoy AI Gateway) integrates with — Agent Router supported-providers page lists DeepSeek. (verified 2026-10-29)
- agentgateway integrates with — agentgateway integrations docs have a DeepSeek provider page. (verified 2026-10-29)
- Bifrost (Maxim AI) integrates with — Bifrost supported-providers docs have a provider page for DeepSeek. (verified 2026-10-29)
- Cloudflare AI Gateway integrates with — Cloudflare AI Gateway docs have a DeepSeek provider page. (verified 2026-10-29)
- Comet ML integrates with — Comet's Opik docs have a DeepSeek R1 / V3 family integration page. (verified 2026-10-29)
- Inspect AI integrates with — Inspect AI model-providers page documents a DeepSeek provider. (verified 2026-10-29)
- Kong AI Gateway integrates with — Kong AI Gateway docs have a DeepSeek AI-provider reference page. (verified 2026-10-29)
- LangChain (framework) integrates with — LangChain docs have a DeepSeek integration provider page (Advanced reasoning and coding AI models). (verified 2026-10-29)
- Langflow integrates with — Langflow components bundle: DeepSeek language model component. (verified 2026-10-29)
- Langfuse integrates with — Langfuse's integrations directory has a DeepSeek R1 / V3 family integration page. (verified 2026-10-29)
- LangSmith integrates with — LangSmith docs have a tracing integration page for DeepSeek R1 / V3 family. (verified 2026-10-29)
- Letta / MemGPT integrates with — Letta models docs list DeepSeek in the supported model provider table. (verified 2026-10-29)
- LiteLLM integrates with — LiteLLM docs have a DeepSeek provider page. (verified 2026-10-29)
- LiveKit Agents integrates with — LiveKit Agents docs have a DeepSeek R1 / V3 family model plugin page. (verified 2026-10-29)
- LlamaIndex (framework) integrates with — LlamaIndex has an official LLM integration package llama-index-llms-deepseek in its repo. (verified 2026-10-29)
- Mem0 integrates with — Mem0 docs have a DeepSeek LLM provider page for Mem0 open source. (verified 2026-10-29)
- Milvus integrates with — Milvus's integrations overview lists a Milvus + DeepSeek R1 / V3 family integration tutorial. (verified 2026-10-29)
- n8n (AI agent flows) integrates with — n8n built-in AI cluster node: DeepSeek Chat Model node. (verified 2026-10-29)
- OpenRouter integrates with — OpenRouter lists DeepSeek as an inference provider (providers API slug "deepseek"). (verified 2026-10-29)
- Phidata (now Agno) integrates with — Agno ships a deepseek model provider module (agno.models.deepseek). (verified 2026-10-29)
- Pipecat (Daily) integrates with — Pipecat docs have a DeepSeek R1 / V3 family service page. (verified 2026-10-29)
- Portkey AI integrates with — Portkey docs have a DeepSeek LLM provider integration page. (verified 2026-10-29)
- Pydantic-AI integrates with — Pydantic AI docs have a DeepSeek model provider page. (verified 2026-10-29)
- Vapi integrates with — Vapi docs list DeepSeek R1 / V3 family as a provider/integration. (verified 2026-10-29)
- Vercel AI Gateway integrates with — Vercel AI Gateway model directory has a DeepSeek V4.1 Flash model page. (verified 2026-10-29)
- Vercel AI SDK integrates with — AI SDK docs have a DeepSeek official AI SDK provider page. (verified 2026-10-29)
- Weaviate integrates with — Weaviate's integrations / model-provider docs list DeepSeek R1 / V3 family as an integration. (verified 2026-10-29)