Model comparison
Is DeepSeek-V4-Pro cheaper than Claude Sonnet 5?
DeepSeek-V4-Pro costs $0.95 for a cached coding session that costs $2.40 on Claude Sonnet 5, and cache writes explain most of it. Rates, limits, and tools.
· Prices as of September 28, 2026
DeepSeek-V4-Pro
DeepSeek · Released August 13, 2026
DeepSeek's agent-focused large model, released in general availability in August 2026 with support for OpenAI's Responses API and Codex.
DeepSeek-V4-Pro facts and comparisonsClaude Sonnet 5
Anthropic · Released June 30, 2026
Anthropic's balance of speed and intelligence, and a drop-in upgrade for Claude Sonnet 4.6. Anthropic pitches it as close to Claude Opus 4.8 at a lower price.
Claude Sonnet 5 facts and comparisons
The short answer
Yes: DeepSeek-V4-Pro costs $0.95 for the example agentic coding session against $2.40 on Claude Sonnet 5, a 2.5x gap, and most of the difference is cache writes, which DeepSeek bills as ordinary input. Claude Sonnet 5 is Anthropic's balance of speed and intelligence and runs in Claude Code, Cursor, and GitHub Copilot. DeepSeek-V4-Pro fits agent work through OpenRouter or OpenCode, or self-hosted under the MIT license, with the caveat that DeepSeek plans a V4.1 Pro successor.
Choose DeepSeek-V4-Pro if
- Your sessions write a lot to the cache: with no write fee from DeepSeek, the example's 400K written tokens cost $0.53 on V4-Pro against $1.30 on Sonnet 5.
- Output is a big share of your spend, and $3.96 per million at DeepSeek's peak beats paying $10.
- You want the weights too, under the MIT license, with DeepSeek-V4-Pro-0813 as the current snapshot.
- Much of your team's day falls outside DeepSeek's peak hours, when every V4-Pro rate drops by 50%.
Choose Claude Sonnet 5 if
- Claude Code is your main tool, and you want the model its sonnet alias points to on the Anthropic API.
- You want pricing that has settled: Sonnet 5's launch price became its standard price on August 10, 2026, and a planned increase was cancelled.
- You are moving up from Claude Sonnet 4.6 and want the model Anthropic pitches as a drop-in upgrade for it.
- You want Anthropic's agentic focus, which it describes as planning and using tools like browsers and terminals on its own.
Side by side
Specs and prices
| Fact | DeepSeek-V4-Pro | Claude Sonnet 5 |
|---|---|---|
| Maker | DeepSeek | Anthropic |
| API model id | deepseek-v4-pro | claude-sonnet-5 |
| Released | August 13, 2026 | June 30, 2026 |
| Status | Current | Current |
| Context window | 1M tokens | 1M tokens |
| Max output | 384K tokens | 128K tokens |
| Open weights | Yes | No |
| Input, per 1M tokens | $1.32 | $2 |
| Cache hit, per 1M | $0.044 | $0.20 |
| Cache write, per 1M | $1.32 (same as input) | $2.50 (5-minute), $4 (1-hour) |
| Output, per 1M tokens | $3.96 | $10 |
| Runs in | OpenCode and OpenRouter | Claude Code, Cursor, OpenCode, OpenRouter, and GitHub Copilot |
Standard API rates in US dollars, as published by each maker (DeepSeek-V4-Pro: September 28, 2026; Claude Sonnet 5: September 26, 2026). Batch and priority tiers, taxes, and subscription plans are not included. DeepSeek-V4-Pro: Prices are DeepSeek's peak rates. Off-peak hours cost 50% less: peak is 01:00 to 04:00 and 06:00 to 10:00 UTC on weekdays, excluding Chinese public holidays. Claude Sonnet 5: The full 1M context window is billed at standard rates. The launch price became the standard price on August 10, 2026, and a planned increase was cancelled.
Cost
What the same work costs
The same token counts, priced at each model’s published rates. Cached tokens are billed at each maker’s cache prices, so the session shows what caching is worth on each model.
Real sessions differ: the two models count the same code as different numbers of tokens, and reasoning settings change how much each one writes. Your own history is the real test.
| Workload | DeepSeek-V4-Pro | Claude Sonnet 5 |
|---|---|---|
| Agentic coding session, 100K input, 400K written to cache, 2M read from cache, 50K output | $0.95 | $2.40 |
| Large one-off review, 150K input with no cache hits, 10K output | $0.24 | $0.40 |
| Output-heavy generation, 30K input, 80K output | $0.36 | $0.86 |
| A month of sessions, 110 sessions: 5 a day, 22 working days | $104.06 | $264.00 |
| Where the session’s cost goes | ||
| Cache writes | $0.53 | $1.30 |
| Cache reads | $0.09 | $0.40 |
| Uncached input | $0.13 | $0.20 |
| Output | $0.20 | $0.50 |
- caching saves on the session with DeepSeek-V4-Pro (73%)
- $2.55
- caching saves on the session with Claude Sonnet 5 (56%)
- $3.10
Cache writes explain most of the 2.5x gap
Per token, DeepSeek-V4-Pro and Claude Sonnet 5 are closer than the session total suggests. Input is $1.32 against $2 per million at DeepSeek's peak rates, 1.5x. That is why the large one-off review, with no cache in play, differs by only $0.16: $0.24 against $0.40.
The example session writes 400K tokens to the cache. DeepSeek lists no fee for writing its cache, so those tokens cost V4-Pro's input rate, $0.53. Anthropic splits the same writes between its 5-minute cache at 1.25x input, $2.50 per million on Sonnet 5, and its 1-hour cache at 2x, $4 per million, for $1.30. That $0.77 difference is more than half of the $1.45 gap on the session.
Reads add $0.31 more, because a V4-Pro cache hit costs $0.044 per million, 3.3% of input, against Sonnet 5's $0.20, the usual 10%. Output adds $0.30, at $3.96 against $10, and fresh input adds $0.07. The session totals $0.95 and $2.40, and over 110 sessions a month $104.06 against $264.00.
How Claude Code's cache setting shifts the comparison
Claude Code decides which Anthropic write price you pay. With an API key its main conversation uses the 5-minute cache, and on a Claude subscription the 1-hour cache. The example session splits its writes evenly between the two, a middle case. An API-key setup would push Sonnet 5's write cost below $1.30, and a subscription above it, though subscriptions are priced differently from API rates anyway.
By Anthropic's own rule of thumb, Sonnet 5 earns back a 5-minute write with one cache read and a 1-hour write with two. DeepSeek's cache has no write premium to earn back, and its disk cache is on by default for every account with no code changes.
DeepSeek's side has its own variables. Its peak rates, used here, apply from 01:00 to 04:00 and 06:00 to 10:00 UTC on weekdays, excluding Chinese public holidays, and other hours cost 50% less, so off-peak V4-Pro sessions widen the gap. The tables also use DeepSeek's own API price, and OpenRouter providers serving the open weights can charge differently.
What changes if you switch from Sonnet 5 to V4-Pro
Anthropic says "Claude Sonnet 5 is built to be the most agentic Sonnet model yet," with performance close to Claude Opus 4.8 at lower prices. Adaptive thinking is on by default at high effort. It runs in Claude Code, Anthropic's own agent, and is also offered in Cursor, OpenRouter, OpenCode, and GitHub Copilot.
DeepSeek positions V4-Pro as its agent-focused large model and says the general availability release "greatly enhances agent capabilities, with particularly significant performance improvements in production environments." It lists reasoning effort at low, high, and max, and native support for OpenAI's Responses API. It runs through OpenRouter and OpenCode, and is not in Cursor or Copilot.
Two DeepSeek statements matter for timing. V4-Pro service continues with billing unchanged until a V4.1 Pro model arrives, and DeepSeek says its newer DeepSeek-V4.1-Flash is ahead of V4-Pro on performance, cost, speed, and total runtime in tests by several parties. Anyone weighing V4-Pro against Sonnet 5 on price may want to price V4.1 Flash as well.
Both models accept 1M tokens of context, and DeepSeek lists 384K of output against 128K. Their tokenizers differ, and Sonnet 5's counts about 1.0 to 1.35x as many tokens as Claude Sonnet 4.6 for the same text, so compare real sessions. EveryToken prices Sonnet 5 from Claude Code at Anthropic's rates and V4-Pro when it runs through OpenRouter.
Prompt caching
How each maker bills cached tokens
DeepSeek
DeepSeek's disk cache is on by default for every account, with no code changes. DeepSeek lists no fee for writing the cache.
A cache hit costs $0.006 per million tokens on DeepSeek-V4.1-Flash and $0.044 on DeepSeek-V4-Pro at peak rates, and off-peak hours cost 50% less.
Source: DeepSeek API: Context caching
Anthropic
Claude caches a prompt prefix up to a breakpoint. One top-level cache_control field places the breakpoint automatically and moves it as the conversation grows, or you can mark up to 4 blocks yourself. Claude Code manages caching for you.
A 5-minute cache write costs 1.25x the input price and a 1-hour write costs 2x. A cache hit costs 0.1x input on most models, 0.05x on Claude Opus 5.5, and 0.025x on Claude Fable 5.1 and Claude Mythos 5.1. Every hit restarts the cache lifetime at no charge.
Anthropic's rule of thumb: a 5-minute write pays for itself after one cache read, and a 1-hour write after two.
In Claude Code, the main conversation uses the 1-hour cache on a Claude subscription and the 5-minute cache with an API key. Each model has its own cache, so switching models starts over.
Source: Anthropic: Prompt caching
Your own numbers
See what DeepSeek-V4-Pro and Claude Sonnet 5 really cost you.
everyaitoken reads your OpenRouter, Claude Code, Cursor, and OpenCode history on your Mac and prices every request at API rates, with what caching saved or cost. $9 once.
FAQ
Questions
How much would DeepSeek-V4-Pro save over Claude Sonnet 5 in a month?
At 110 example sessions a month, $104.06 on DeepSeek-V4-Pro against $264.00 on Claude Sonnet 5, a $159.94 difference at DeepSeek's peak rates. Those are API-equivalent estimates, not what a subscription plan charges.
Can Claude Code run DeepSeek-V4-Pro?
Claude Code, Anthropic's own coding agent, is built around Claude models and defaults to them. Custom configuration can point it at other providers' compatible endpoints, a setup outside the scope of this page. DeepSeek-V4-Pro runs through OpenCode and OpenRouter, and DeepSeek says it also supports OpenAI's Responses API natively.
Why is the uncached gap smaller than the session gap?
Without the cache, only list rates count: $1.32 against $2 for input, so a large review differs by 1.7x. The session adds cache writes and reads, where DeepSeek's fee-free writes and $0.044 hits pull further ahead, for 2.5x.
Is a newer DeepSeek Pro model coming?
DeepSeek has told V4-Pro users that the service keeps running, with billing unchanged, until a V4.1 Pro model arrives. No V4.1 Pro model is part of this comparison.
Sources
- DeepSeek API: Models and pricing
- DeepSeek API: Change log
- DeepSeek: V4 Pro release
- OpenRouter: DeepSeek-V4-Pro-0813
- OpenCode docs: Zen
- Anthropic: Pricing
- Anthropic docs: Claude Sonnet 5
- Anthropic: Introducing Claude Sonnet 5
- Claude Code docs: Model configuration
- Cursor docs: Claude Sonnet 5
- OpenRouter: Claude Sonnet 5
- OpenCode docs: Zen
- GitHub Docs: Supported AI models in Copilot
- DeepSeek API: Context caching
- Anthropic: Prompt caching