{"slug": "deepseek-v4-flash-0731-benchmarks-pricing-and-dev-setup-guide", "title": "DeepSeek V4 Flash 0731: Benchmarks, Pricing, and Dev Setup Guide", "summary": "DeepSeek released V4-Flash-0731 on July 31, 2025, and the smaller model now outperforms the larger V4-Pro-Preview across all nine published agentic benchmarks, including Terminal Bench 2.1 (82.7 vs. 72.1) and DeepSWE (54.4 vs. 12.8), at one-third the output cost. The model retains the same 284B MoE architecture, 13B active parameters, and 1M context window, with native Responses API support added. DeepSeek also retired the 'deepseek-chat' and 'deepseek-reasoner' aliases on July 24, requiring migration to 'deepseek-v4-flash'.", "body_md": "DeepSeek made V4 Flash official on July 31, and the buried headline is this: the smaller model now beats the larger one. V4-Flash-0731 outperforms V4-Pro-Preview across all nine published agentic benchmarks — [according to DeepSeek’s own release notes](https://deepseek.ai/blog/deepseek-v4-flash-ga-agent-benchmarks). Terminal Bench 2.1 sits at 82.7 for Flash versus Pro’s 72.1. DeepSWE lands at 54.4 versus 12.8. Cybergym: 76.7 versus 52.7. The smaller model wins every category, at a third of the output cost. If you’re running coding agents and haven’t switched yet, here’s your migration guide.\n\n## What Changed From Preview to GA\n\nAlmost nothing architectural changed — and that’s the interesting part. DeepSeek re-ran only the post-training phase. Same 284B MoE architecture, same 13B active parameters per token, same 1M context window. The model string stays `deepseek-v4-flash`\n\n, so existing integrations require no code changes. What did land in this build: native Responses API support, which makes it directly compatible with Codex-based integrations out of the box.\n\nOne housekeeping note before setup: if you’re still calling `deepseek-chat`\n\nor `deepseek-reasoner`\n\n, those aliases **retired on July 24**. Switch to `deepseek-v4-flash`\n\nnow or your calls are failing silently.\n\n## The Benchmark Numbers Worth Knowing\n\n| Benchmark | Flash-0731 | Flash Preview | Pro Preview |\n|---|---|---|---|\n| Terminal Bench 2.1 | 82.7 | 61.8 | 72.1 |\n| DeepSWE | 54.4 | 7.3 | 12.8 |\n| Cybergym | 76.7 | 38.7 | 52.7 |\n| NL2Repo | 54.2 | 39.4 | 38.5 |\n| Toolathlon (verified) | 70.3 | 49.7 | 55.9 |\n\nOn DeepSWE — a real-world software engineering benchmark — Flash is more than four times better than its preview version and four times better than Pro Preview. Independent verification from Artificial Analysis puts Flash at #2 of 162 models measured. One developer in the [HackerNews discussion](https://news.ycombinator.com/item?id=49214008) put it plainly: “It would absolutely have been a frontier model last December.”\n\n## What It Costs — Honestly\n\n| Token Type | V4-Flash | V4-Pro |\n|---|---|---|\n| Input (cache miss) | $0.14/M | $0.435/M |\n| Input (cache hit) | $0.0028/M | $0.003625/M |\n| Output | $0.28/M | $0.87/M |\n\nOutput tokens cost 3x less than Pro. Cache hits knock 98% off input costs — critical for agentic loops where your system prompt repeats on every tool call. In practice, a full day of agent-heavy work runs under $0.50 on Flash versus roughly $2 on Pro.\n\nThere’s a real caveat here: Flash generates approximately three times the median token volume on agentic tasks due to its internal reasoning chain. Effective cost per completed task runs higher than raw rates suggest — but it’s still cheaper than Pro in every real-world scenario. Also worth noting: DeepSeek has announced a **2x peak-hour surcharge** during Beijing business hours, effective date still TBD.\n\n## Setting Up in Claude Code\n\nDeepSeek provides an Anthropic-compatible endpoint, so setup is a handful of environment variables. The [official integration docs](https://api-docs.deepseek.com/quick_start/agent_integrations/claude_code/) cover it, but the smart mapping is to assign Flash to lighter slots (Haiku, subagents) and Pro to the main model role:\n\n```\nexport ANTHROPIC_BASE_URL=https://api.deepseek.com/anthropic\nexport ANTHROPIC_AUTH_TOKEN=your_deepseek_api_key\nexport ANTHROPIC_MODEL=deepseek-v4-pro\nexport ANTHROPIC_DEFAULT_OPUS_MODEL=deepseek-v4-pro\nexport ANTHROPIC_DEFAULT_SONNET_MODEL=deepseek-v4-pro\nexport ANTHROPIC_DEFAULT_HAIKU_MODEL=deepseek-v4-flash\nexport CLAUDE_CODE_SUBAGENT_MODEL=deepseek-v4-flash\nexport CLAUDE_CODE_EFFORT_LEVEL=max\nexport CLAUDE_CODE_AUTO_COMPACT_WINDOW=786432\n```\n\nThis maps Flash to the high-frequency sub-tasks and reserves Pro for the primary reasoning pass. Given that Flash outperforms Pro on every agentic benchmark, you could map everything to Flash — but the hybrid approach is the sensible starting point while Pro’s GA is still pending.\n\n## Setting Up in Cursor\n\nOpen Settings (Cmd+, on Mac, Ctrl+, on Windows), go to Models, and toggle “Override OpenAI Base URL.” Enter `https://api.deepseek.com`\n\n— no trailing `/v1`\n\n. Paste your DeepSeek API key into the OpenAI key field, click “+ Add model,” type `deepseek-v4-flash`\n\n, and hit Verify. Switch to it in the chat model picker before starting a session.\n\nThere’s one important caveat here: Cursor’s Composer panel has a known issue with extended tool-call sequences. Cursor only forwards the `content`\n\nfield to the model — not `reasoning_content`\n\n— which causes failures when DeepSeek’s thinking mode needs to replay its chain across multiple requests. Use the Chat panel for complex multi-step tasks or disable thinking mode for Composer sessions. Background Agents also don’t support custom models as of this writing.\n\n## Get Your API Key\n\nIf you don’t have one, start at the [DeepSeek platform](https://platform.deepseek.com/api_keys). Minimum wallet funding is $5 — that covers weeks of normal agent usage at Flash pricing.\n\n## The Takeaway\n\nV4-Flash is the agent model of summer 2026. A frontier-adjacent model at sub-commodity pricing, MIT-licensed, open weights available for self-hosting on 4x A100s. V4-Pro still hasn’t hit GA — when it does, the picture may shift. For now, Flash is the default choice for coding agents. The benchmarks say so, and so does your wallet.", "url": "https://wpnews.pro/news/deepseek-v4-flash-0731-benchmarks-pricing-and-dev-setup-guide", "canonical_source": "https://byteiota.com/deepseek-v4-flash-0731-benchmarks-pricing-dev-setup/", "published_at": "2026-08-10 06:09:15+00:00", "updated_at": "2026-08-10 06:35:01.494939+00:00", "lang": "en", "topics": ["artificial-intelligence", "large-language-models", "ai-products", "ai-tools", "ai-research"], "entities": ["DeepSeek", "V4-Flash-0731", "V4-Pro-Preview", "Terminal Bench 2.1", "DeepSWE", "Cybergym", "Artificial Analysis", "Claude Code"], "alternates": {"html": "https://wpnews.pro/news/deepseek-v4-flash-0731-benchmarks-pricing-and-dev-setup-guide", "markdown": "https://wpnews.pro/news/deepseek-v4-flash-0731-benchmarks-pricing-and-dev-setup-guide.md", "text": "https://wpnews.pro/news/deepseek-v4-flash-0731-benchmarks-pricing-and-dev-setup-guide.txt", "jsonld": "https://wpnews.pro/news/deepseek-v4-flash-0731-benchmarks-pricing-and-dev-setup-guide.jsonld"}}