What other models does DeepSeek offer?

DeepSeek builds an open-weight model family that spans cost-efficient non-reasoning models and reasoning-capable models, with the April 2026 V4 generation representing the current generation at both quality tiers. Here are some other models DeepSeek supports:

• DeepSeek V4 Flash: The cost-optimized reasoning model released alongside V4 Pro in April 2026, built on a 284B-parameter MoE architecture with 13B active parameters and a 1M-token context window. It ranks #11 of 89 comparable reasoning models on the Artificial Analysis Intelligence Index (as of 06/08/2026), generates 112.9 tokens per second, and is priced at $0.14 input / $0.28 output per 1M tokens, making it the lowest-cost reasoning option in DeepSeek's lineup

• DeepSeek V3.2: The current top-tier non-reasoning model in DeepSeek's V3 family, released December 2025, with 685B total parameters, 37B active, and a 128k-token context window. It scores #12 of 43 comparable non-reasoning models on the Intelligence Index (as of 06/08/2026) and costs $0.50 input / $1.60 output per 1M tokens

• DeepSeek V3: The original December 2024 non-reasoning model that established DeepSeek's open-weight presence, with 671B total parameters and a 128k context window. It's available at $0.40 input / $0.89 output per 1M tokens for teams already on that API surface

• DeepSeek R1: DeepSeek's earlier reasoning model with a 128k-token context window and chain-of-thought capability, priced at $1.35 input / $4.20 output per 1M tokens. It's significantly more expensive than V4 Pro for a lower Intelligence Index score, making it relevant mainly for teams with existing R1 integrations

How does DeepSeek V4 Pro differ from DeepSeek's other models?

DeepSeek V4 Pro is the most capable model in DeepSeek's current lineup, sitting at the top of the reasoning tier with the highest Intelligence Index score, the largest parameter count, and the longest context window in the family.

• Intelligence ranking: V4 Pro scores #3 of 89 comparable reasoning models on the Artificial Analysis Intelligence Index (as of 06/08/2026), with an index score of 52. DeepSeek V4 Flash, the next model down, scores 47 and ranks #11. The gap of 5 index points translates to meaningfully better performance on hard reasoning tasks including scientific analysis, complex coding, and long-chain logic

• Pricing: At $0.435 per 1M input tokens and $0.87 per 1M output tokens, V4 Pro costs approximately 3x more per input token than V4 Flash ($0.14) (as of 06/08/2026). The cache hit rate is 99%, bringing repeated-context costs to $0.004 per 1M tokens. For budget-sensitive workloads, V4 Flash covers most reasoning tasks at a fraction of the price

• Architecture: V4 Pro is built on 1.6 trillion total parameters with 49B active in a Mixture of Experts configuration. V4 Flash uses 284B total with 13B active. The substantially larger base gives V4 Pro more model capacity for nuanced multi-step reasoning

• Speed: V4 Pro generates 61.6 output tokens per second (as of 06/08/2026), ranking #28 of 89 comparable models. V4 Flash generates 112.9 tokens per second, nearly double the throughput. For latency-sensitive applications, V4 Flash is the faster option within the same provider

• Context window: Both V4 Pro and V4 Flash support 1M-token context windows. This is a major shared advantage over the V3 series, which is capped at 128k tokens across all variants

DeepSeek V4 Pro is the right choice when task difficulty justifies its cost premium over V4 Flash. It's best suited for production workloads involving hard scientific reasoning, complex multi-file code generation, or agentic pipelines where a wrong answer has a meaningful downstream cost.

What models should I consider using alongside DeepSeek V4 Pro?

No single model is optimal for every task. Here are models worth pairing with DeepSeek V4 Pro depending on what your product needs:

• DeepSeek V4 Flash (DeepSeek): For the majority of reasoning requests in your pipeline that don't require V4 Pro's full intelligence ceiling, route to V4 Flash. At roughly one-third the input cost and nearly double the output speed, it handles most reasoning workloads well and makes V4 Pro available as a targeted escalation path for the hardest tasks

• Claude Sonnet 4.6 (Anthropic): For instruction-following tasks, enterprise document workflows, or use cases that require a western-hosted provider for compliance reasons, Claude Sonnet 4.6 delivers strong, consistent performance and is a natural cross-provider complement to DeepSeek V4 Pro

• Gemini 2.5 Pro (Google): For tasks that involve image or document inputs alongside text, Gemini 2.5 Pro adds multimodal capability that V4 Pro lacks entirely. It's a strong pairing when your pipeline includes mixed-modality requests

• GPT-4o mini (OpenAI): For high-volume, low-complexity requests like classification, summarization, or short-form extraction where reasoning depth isn't required, GPT-4o mini handles the simple end of the request distribution at low cost, freeing V4 Pro's budget for harder problems

• Llama 3.3 70B (Meta): For non-reasoning workloads where self-hosted or on-premises deployment is a requirement, Llama 3.3 70B provides a capable open-weight fallback that can cover general-purpose generation tasks without incurring API costs

What are the challenges of using DeepSeek V4 Pro in my product?

Like any production LLM, DeepSeek V4 Pro comes with tradeoffs worth planning for:

• Cost at scale: At $0.87 per 1M output tokens, costs compound quickly at high request volumes. Reasoning models also generate significantly more output tokens than non-reasoning models due to chain-of-thought traces, meaning the effective cost per user query is often higher than the nominal rate suggests. Without active budget tracking, production spend can escalate unexpectedly

• Output speed: At 61.6 tokens per second (as of 06/08/2026), V4 Pro is slower than both V4 Flash (112.9 t/s) and many competing models. For streaming interfaces where perceived response time matters to end users, this latency is a concrete user experience constraint

• Text-only modality: V4 Pro accepts and outputs text only. Any pipeline requiring image, audio, or video understanding must route those request types to a different model, which adds routing logic and potential latency for multimodal workflows

• Provider dependency: Concentrating high-value inference on DeepSeek as a single provider creates fragility during outages or if the V4 Pro checkpoint is deprecated. Because V4 Pro has no direct equivalent elsewhere in terms of parameter scale and open-weight availability, failover planning requires careful selection of a comparable closed-weights alternative

• New model track record: Released in April 2026, V4 Pro has a relatively short production history. Edge case behavior, prompt engineering best practices, and community-documented reliability patterns are still developing compared to more established models

Why should I use Merge Gateway to route LLM requests with DeepSeek V4 Pro and every other model?

Using DeepSeek V4 Pro through Merge Gateway gives you access to the model itself and the infrastructure layer around it:

• Build Your Own Router: Define what "best" means for your traffic by selecting from curated ML benchmarks or adding your own eval scores. For a premium model like V4 Pro, this is especially valuable: the router can automatically direct only the requests that justify V4 Pro's cost while falling back to V4 Flash or another model for the rest, with a plain-language explanation of every decision

• One API, every provider: Access DeepSeek V4 Pro and every other major LLM through a single endpoint and API key. Change providers by swapping the model string, with no application code changes required

• Intelligent routing and automatic failover: Merge routes around DeepSeek outages automatically. Routing policies based on cost, latency, or quality can reduce spend by 40–60% without touching your application code

• Cost governance: Set hard or soft project budgets so DeepSeek V4 Pro spend stays within plan. Every request is attributed to a model, project, and tag in a unified billing dashboard across all providers

• Security and compliance controls: Apply DLP rules and prompt injection protection before every request reaches DeepSeek. Enforce per-project model and region policies without adding that logic to your application

How can I start routing requests to DeepSeek V4 Pro via Merge Gateway?

Getting DeepSeek V4 Pro running through Merge Gateway takes a few minutes:

1. Create an account and get your API key from the dashboard.

2. Install the Merge Gateway SDK: run pip install merge-gateway-sdk (Python) or npm install merge-gateway-sdk (Node). Alternatively, if you're already using the OpenAI SDK, set base_url = "https://api-gateway.merge.dev/v1/openai" and your existing code works as-is.

3. Make your first request using the provider/model format. For DeepSeek V4 Pro, the model string is deepseek/deepseek-v4-pro. Swap the model string to route to any other provider without changing anything else.

4. Configure a routing policy in the dashboard to set failover behavior, cost limits, and optimization strategy. Your first policy can be as simple as naming DeepSeek V4 Pro as primary with DeepSeek V4 Flash or a cross-provider reasoning model as fallback.

Full setup instructions and SDK references are in the Merge Gateway docs.