DeepSeek V4 Flash (0731)
DeepSeekhuggingface.co/deepseek-ai/DeepSeek-V4-Flash-073
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....
Updated 3 h ago · first seen 11 Sept 2026
model_01M294AJ1F4RJVP5FJ0MN8KYAB
- Parameters
- 304.2B
- T2 · 4 h ago
- Context
- 1.31M tokens
- T2 · 5 h ago
- Released
- 31 Jul 2026
- T2 · 5 h ago
- License
- mit
- T2 · 4 h ago
Specification
- Release date
- 31 Jul 2026
Source:OpenRouter public model & pricing listingT2observed 5 h agomedium
- Openness
- open-weights
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 4 h agomedium
- License
- mit
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 4 h agomedium
- Architecture
- DeepseekV4ForCausalLM
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 4 h agomedium
- Parameters
- 304.2B
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 4 h agomedium
- Context window
- 1.31M tokens
Source:OpenRouter public model & pricing listingT2observed 5 h agomedium
- Max output
- 943.7K tokens
Source:OpenRouter public model & pricing listingT2observed 5 h agomedium
- Modalities
- text
Source:OpenRouter public model & pricing listingT2observed 5 h agomedium
- Input modalities
- text
Source:OpenRouter public model & pricing listingT2observed 5 h agomedium
- Output modalities
- text
Source:OpenRouter public model & pricing listingT2observed 5 h agomedium
- Tokenizer
- DeepSeek
Source:OpenRouter public model & pricing listingT2observed 5 h agomedium
- Quantization format
- fp8
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 4 h agomedium
- File size
- 166.9 GB
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 4 h agomedium
- Hugging Face repo
- deepseek-ai/DeepSeek-V4-Flash-0731
Source:OpenRouter public model & pricing listingT2observed 5 h agomedium
- Pipeline tag
- text-generation
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 4 h agomedium
- Model card
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 4 h agomedium
- Downloads
- 4,397,122
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 3 h agomedium
- Likes
- 3,931
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 3 h agomedium
- Gated
- No
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 4 h agomedium
- Hf inference providers
- baseten, deepinfra, fireworks-ai, novita, scaleway, together
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 4 h agomedium
- Is quantized
- Yes
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 4 h agomedium
- Last modified
- 2026-08-01T03:07:41+00:00
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 4 h agomedium
- Library name
- transformers
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 4 h agomedium
- Livebench hf link
- https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash-0731
Source:LiveBenchT2observed 3 h agomedium
- Downloads all time
- 5,938,643
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 3 h agomedium
- Model type
- deepseek_v4
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 4 h agomedium
- Openrouter id
- deepseek/deepseek-v4-flash-0731
Source:OpenRouter public model & pricing listingT2observed 5 h agomedium
- Supported parameters
- frequency_penalty, include_reasoning, logit_bias, logprobs, max_tokens, min_p, parallel_tool_calls, presence_penalty, reasoning, reasoning_effort, repetition_penalty, response_format, seed, stop, structured_outputs, temperature, tool_choice, tools, top_a, top_k, top_logprobs, top_p
Source:OpenRouter public model & pricing listingT2observed 5 h agomedium
- Tags
- transformers, safetensors, deepseek_v4, text-generation, 8-bit, fp8
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 4 h agomedium
- Weights dtype
- BF16, F32, F8_E4M3, I64, I8
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 4 h agomedium
Each value shows its source, tier and observation time. Conflicting claims are kept side by side and flagged — never averaged. How AI Atlas records facts →
Provenance
Attributed facts
34
Source tiers
T234
Freshest observation
3 h ago
Conflicts
None
Modalities
- Modalities
- text
- Input
- text
- Output
- text
Capabilities
Tool calling
Yes
OpenRouter public model & pricing listing · T2
Structured output
Yes
OpenRouter public model & pricing listing · T2
Reasoning
Yes
OpenRouter public model & pricing listing · T2
Vision
Unavailable
Audio
Unavailable
Fine-tuning available
Unavailable
- Context window
- 1.31M tokens
Source:OpenRouter public model & pricing listingT2observed 5 h agomedium
- Max output
- 943.7K tokens
Source:OpenRouter public model & pricing listingT2observed 5 h agomedium
- Tokenizer
- DeepSeek
Source:OpenRouter public model & pricing listingT2observed 5 h agomedium
| Benchmark | Score | Metric | Config | Evaluated | Source |
|---|---|---|---|---|---|
| LiveBench | 74.17% | global_average | release=2026-06-25 · aggregation=mean of category averages; category = mean of its subtasks · livebench_model_id=deepseek-v4-flash-0731 | 25 Jun 2026 | livebench.aiT2 |
| LiveBench | 86.64% | category:Reasoning | release=2026-06-25 · subtasks=["theory_of_mind","zebra_puzzle","spatial","logic_with_navigation"] · livebench_model_id=deepseek-v4-flash-0731 | 25 Jun 2026 | livebench.aiT2 |
| LiveBench | 74.98% | category:Coding | release=2026-06-25 · subtasks=["code_generation","code_completion"] · livebench_model_id=deepseek-v4-flash-0731 | 25 Jun 2026 | livebench.aiT2 |
| LiveBench | 46.77% | category:Agentic Coding | release=2026-06-25 · subtasks=["javascript","typescript","python"] · livebench_model_id=deepseek-v4-flash-0731 | 25 Jun 2026 | livebench.aiT2 |
| LiveBench | 86.79% | category:Mathematics | release=2026-06-25 · subtasks=["AMPS_Hard","integrals_with_game","math_comp","olympiad"] · livebench_model_id=deepseek-v4-flash-0731 | 25 Jun 2026 | livebench.aiT2 |
| LiveBench | 79.33% | category:Data Analysis | release=2026-06-25 · subtasks=["consecutive_events","tablejoin","tablereformat"] · livebench_model_id=deepseek-v4-flash-0731 | 25 Jun 2026 | livebench.aiT2 |
| LiveBench | 79.18% | category:Language | release=2026-06-25 · subtasks=["connections","plot_unscrambling","typos"] · livebench_model_id=deepseek-v4-flash-0731 | 25 Jun 2026 | livebench.aiT2 |
| LiveBench | 65.52% | category:IF | release=2026-06-25 · subtasks=["paraphrase","simplify","story_generation","summarize"] · livebench_model_id=deepseek-v4-flash-0731 | 25 Jun 2026 | livebench.aiT2 |
Scores are reported as published, with their evaluation configuration (harness, prompting, judge). Results with different configs are not directly comparable — see methodology.
Current prices
| Provider | Input / 1M | Output / 1M | Cached in | Batch in / out | Context | Observed | Source |
|---|---|---|---|---|---|---|---|
| DeepSeek APIdeepseek/deepseek-v4-flash-0731 | $0.065 | $0.18 | $0.016 | — | 1.05M | 32 min ago | openrouter.aiT2 |
| DeepSeek APIdeepseek/deepseek-v4-flash-0731:batch | $0.11 | $0.33 | $0.0035 | — | 1.05M | 32 min ago | openrouter.aiT2 |
| Together AIdeepseek-v4-flash-0731 | $0.14 | $0.28 | $0.03 | — | — | 3 h ago | together.aiT1 |
| Fireworks AIfireworks/deepseek-v4-flash-0731 | $0.22 | $0.66 | $0.007 | — | — | 5 h ago | app.fireworks.aiT1 |
USD per 1M tokens as published by each provider (USD). Rows are append-only: every change is kept in the history below.
Available through 3
Price history
Input price · USD / 1M tokens
Output price · USD / 1M tokens
- Fireworks AI
- DeepSeek API
- Together AI
Memory need = bytes per parameter (4-bit ≈ 0.5 × 1.15 overhead, 8-bit 1.0, fp16 2.0) + a KV-cache allowance. Not a measurement.
| Hardware | Quantization | Memory | Est. need | Fits |
|---|---|---|---|---|
| NVIDIA DGX B200 | 4bit | 1,440 GB | 175.4 GB | Yes |
| Apple M3 Ultra | 4bit | — | 175.4 GB | Yes |
| Mac Studio (Apple M5 Ultra) | 4bit | — | 175.4 GB | Yes |
| AMD Instinct MI325X | 4bit | 256 GB | 175.4 GB | Yes |
| AMD Instinct MI300X | 4bit | 192 GB | 175.4 GB | Yes |
| Apple M2 Ultra | 4bit | — | 175.4 GB | Yes |
| NVIDIA B200 | 4bit | 180 GB | 175.4 GB | Yes |
| NVIDIA H200 | 4bit | 141 GB | 175.4 GB | No |
| NVIDIA H200 NVL | 4bit | 141 GB | 175.4 GB | No |
| Apple M1 Ultra | 4bit | — | 175.4 GB | No |
| Apple M3 Max | 4bit | — | 175.4 GB | No |
| Apple M4 Max | 4bit | — | 175.4 GB | No |
| Mac Studio (Apple M5 Max) | 4bit | — | 175.4 GB | No |
| MacBook Pro (Apple M5 Max) | 4bit | — | 175.4 GB | No |
| NVIDIA DGX Spark | 4bit | 128 GB | 175.4 GB | No |
| Apple M2 Max | 4bit | — | 175.4 GB | No |
| NVIDIA H100 NVL | 4bit | 94 GB | 175.4 GB | No |
| NVIDIA A100 80GB | 4bit | 80 GB | 175.4 GB | No |
| NVIDIA H100 SXM | 4bit | 80 GB | 175.4 GB | No |
| Apple M1 Max | 4bit | — | 175.4 GB | No |
| Apple M4 Pro | 4bit | — | 175.4 GB | No |
| Mac mini (Apple M5 Pro) | 4bit | — | 175.4 GB | No |
| MacBook Pro (Apple M5 Pro) | 4bit | — | 175.4 GB | No |
| Apple M3 Pro | 4bit | — | 175.4 GB | No |
| Apple M1 Pro | 4bit | — | 175.4 GB | No |
| Apple M2 Pro | 4bit | — | 175.4 GB | No |
| Apple M4 | 4bit | — | 175.4 GB | No |
| iMac (Apple M4) | 4bit | — | 175.4 GB | No |
| Mac mini (Apple M6) | 4bit | — | 175.4 GB | No |
| MacBook Air (Apple M5) | 4bit | — | 175.4 GB | No |
| MacBook Pro (Apple M5) | 4bit | — | 175.4 GB | No |
| NVIDIA GeForce RTX 5090 | 4bit | 32 GB | 175.4 GB | No |
| Apple M2 | 4bit | — | 175.4 GB | No |
| Apple M3 | 4bit | — | 175.4 GB | No |
| NVIDIA GeForce RTX 3090 | 4bit | 24 GB | 175.4 GB | No |
| NVIDIA GeForce RTX 4090 | 4bit | 24 GB | 175.4 GB | No |
| Apple M1 | 4bit | — | 175.4 GB | No |
No lineage recorded
Papers 1
- arXiv:2606.19348Active35
Repositories 0
No repositories linked yet.
As of
Rewind the record: see this entity's attributes exactly as AI Atlas knew them on a given day.
Claim history · File size
File sizefile_size_gb1
| Value | Valid from → to | Status | Source | Confidence | Extractor |
|---|---|---|---|---|---|
| 166.9 GB | → current | current | Hugging Face Hub (public pages, model cards, papers)T2 | medium | deterministic |
Claims are temporal and append-only: a new observation closes the previous claim (valid_to) instead of overwriting it. Conflicting claims from different sources are kept side by side and flagged — never averaged. Methodology →
DeepSeek V4 Flash (0731) scores 65.517% on LiveBench
livebench_leaderboardDeepSeek V4 Flash (0731) scores 79.175% on LiveBench
livebench_leaderboardDeepSeek V4 Flash (0731) scores 79.327% on LiveBench
livebench_leaderboardDeepSeek V4 Flash (0731) scores 86.79% on LiveBench
livebench_leaderboardDeepSeek V4 Flash (0731) scores 46.768% on LiveBench
livebench_leaderboardDeepSeek V4 Flash (0731) scores 74.984% on LiveBench
livebench_leaderboardDeepSeek V4 Flash (0731) scores 86.635% on LiveBench
livebench_leaderboardDeepSeek V4 Flash (0731) scores 74.171% on LiveBench
livebench_leaderboardTogether AI lists DeepSeek V4 Flash (0731) at $0.14 in / $0.28 out per 1M tokens
together_pricingDeepSeek API lists DeepSeek V4 Flash (0731) at $0.11 in / $0.33 out per 1M tokens
openrouterDeepSeek API lists DeepSeek V4 Flash (0731) at $0.065 in / $0.18 out per 1M tokens
openrouterNew model: DeepSeek V4 Flash (0731) (DeepSeek)
fireworks_pricingFireworks AI lists DeepSeek V4 Flash (0731) at $0.22 in / $0.66 out per 1M tokens
fireworks_pricing
| Source | Document | Type | Tier | Last observed | Snapshots |
|---|---|---|---|---|---|
| Fireworks AI — pricing | docs.fireworks.ai/serverless/pricing.md | pricing | T1· Official | 3 h ago | 1 |
| Together AI — pricing | together.ai/pricing | pricing | T1· Official | 3 h ago | 1 |
| OpenRouter public model & pricing listing | openrouter.ai/api/v1/models | catalogue | T2· Quality secondary | 32 min ago | 4 |
| Hugging Face Hub (public pages, model cards, papers) | huggingface.co/deepseek-ai/DeepSeek-V4-Flash-0731/raw/main/README.md | model_card | T2· Quality secondary | 2 h ago | 1 |
| Hugging Face Hub (public pages, model cards, papers) | huggingface.co/deepseek-ai/DeepSeek-V4-Flash-0731 | model_page | T2· Quality secondary | 3 h ago | 2 |
| Hugging Face Hub (public pages, model cards, papers) | huggingface.co/models?author=deepseek-ai&p=0&sort=downloads | listing | T2· Quality secondary | 3 h ago | 3 |
| LiveBench | livebench.ai/table_2026_06_25.csv | leaderboard | T2· Quality secondary | 3 h ago | 1 |
Tier 1 = official/primary, 2 = quality secondary, 3 = community, 4 = unverified. Every snapshot is archived; see all sources and the methodology.