DeepSeek V4 Flash 0731 is a sparse mixture-of-experts model from DeepSeek, with 13B active parameters out of 284B total. This re-post-trained revision is suited for coding, reasoning, and agent workflows. This is the GA release of DeepSeek V4 Flash.
Modalities
In / Out Price
$0.04998 / $0.09996per 1M
Context
1M
Released
Jul 31, 2026
Different companies host the same model. OpenRouter routes your request to one of them based on the routing mode you pick — Balanced (price + speed), Nitro (fastest), or Exacto (highest tool-calling accuracy).
The average price customers actually pay for this model, next to the prices providers post. Caching and discounts mean the price actually paid is often well below the listed one.
Throughput is how fast the model writes (tokens per second — higher is better). Latency is total round-trip time (lower is better). TTFT is time-to-first-token — how long before you see anything appear (lower is better).
Uptime is the percentage of the past 3 days that at least one provider was responding to requests. Availability is the percentage of time that inference was successfully served. OpenRouter continuously monitors and uses the next-best provider when one returns an error.
Scores on standardized evaluations. Higher percentages are better — and rank percentile shows where this model lands among all models on OpenRouter.
Public apps that send the most traffic to this model. Good signal for what real production workloads look like — and a hint at which use cases this model is best suited for.
Token volume and request traffic to this model over time.
Drop-in code to call this model. OpenRouter's API is OpenAI-compatible — most SDKs work by just swapping the base URL. The only thing that changes between models is the model slug below.
64% off | $0.14$0.04998 | $0.28$0.09996 | $0.028$0.009996 | 0.60s | 125 tps | |
| $0.05 | $0.16 | $0.013 | 0.58s | 13 tps | ||
| $0.065 | $0.18 | $0.016 | 0.68s | 145 tps | ||
| $0.065 | $0.18 | $0.02 | 1.03s | 40 tps | ||
| $0.08 | $0.252 | $0.0252 | 0.52s | 33 tps | ||
| $0.08 | $0.18 | $0.016 | 0.65s | 35 tps | ||
80% off | $0.44$0.088 | $1.32$0.264 | $0.014$0.0028 | 0.75s | 155 tps | |
| $0.09 | $0.195 | $0.0196 | 0.51s | 104 tps | ||
| $0.10 | $0.25 | $0.05 | 0.67s | 167 tps | ||
| $0.11 | $0.66 | $0.007 | 0.77s | 153 tps | ||
| $0.1234 | $0.3475 | $0.03125 | 1.06s | 40 tps | ||
| $0.13 | $0.26 | $0.028 | 0.45s | 95 tps | ||
| $0.13 | $0.26 | $0.028 | 0.42s | 79 tps | ||
| $0.13 | $0.28 | $0.03 | 0.63s | 51 tps | ||
| $0.13 | $0.28 | $0.07 | 0.54s | 121 tps | ||
| $0.14 | $0.28 | $0.03 | 0.88s | 33 tps | ||
| $0.14 | $0.28 | $0.05 | 0.70s | 43 tps | ||
| $0.16 | $0.50 | -- | 0.71s | 32 tps | ||
| $0.175 | $0.35 | $0.035 | 0.81s | 59 tps | ||
| $0.176 | $0.528 | $0.0176 | 1.01s | 65 tps | ||
| $0.22 | $0.66 | $0.007 | 0.77s | 88 tps | ||
| $0.22 | $0.66 | $0.028 | 1.48s | 72 tps | ||
| $0.22 | $0.66 | $0.007 | 0.71s | 84 tps | ||
| $0.40 | $1.20 | $0.01 | 1.99s | 76 tps | ||
7% off | $0.44$0.4092 | $1.32$1.228 | $0.028$0.02604 | 1.31s | 89 tps | |
| $0.44 | $1.32 | $0.028 | 0.58s | 41 tps | ||
| $0.44 | $1.32 | $0.028 | 1.80s | 53 tps | ||
| $0.44 | $1.32 | $0.014 | 0.65s | 63 tps | ||
20% off | $0.44$0.352 | $1.32$1.056 | $0.014$0.0112 | 1.74s | 108 tps |
Throughput
167tok/s
P50, best across providers
Latency
0.42s
P50, best provider
100.00%
99.75%
When an error occurs in an upstream provider, we can recover by routing to another healthy provider, if your request filters allow it. You can access per-provider uptime data programmatically through the Endpoints API. Learn more about our load balancing and customization options.
DeepSeek V4 Flash 0731 is a sparse mixture-of-experts model from DeepSeek, with 13B active parameters out of 284B total. This re-post-trained revision is suited for coding, reasoning, and agent workflows. This is the GA release of DeepSeek V4 Flash.
DeepSeek V4 Flash 0731 costs $0.04998/M input tokens and $0.09996/M output tokens, with separate rates for Cache Read at $0.009996/M tokens.
DeepSeek V4 Flash 0731 has a 1,310,720 token context window. It supports up to 131,072 completion tokens.
Yes. DeepSeek V4 Flash 0731 accepts tools and tool_choice for function calling. It also supports structured outputs via a JSON schema in response_format.
DeepSeek V4 Flash 0731 is served by 28 providers on OpenRouter: Baidu Qianfan, OpenInference, Relace, Sail Research, DigitalOcean, DeepInfra, StreamLake, Makora and 20 more. Requests are routed to the best available provider, with automatic failover to the others, and you can pin or exclude providers with provider routing.
DeepSeek V4 Flash 0731 was released on July 31, 2026.