DeepSeek vs Claude: coding, reasoning and price

DeepSeek V4.1 Flash 1017 costs $0.0042 per example request on Novita AI versus $0.0600 for Claude Opus 5.5, and its AA output speed is 169.0 versus 96.8 tokens/s; choose Claude for agentic coding, where it leads Terminal-Bench 4.0 by 25.8 pp (59.6% vs 33.8%).

Updated · 4 sources · Methodology

DimensionWinnerMarginEvidence
Coding benchmarkClaude Opus 5.525.8 ppTerminal-Bench 4.0: Claude 59.6% vs DeepSeek 33.8%12
Reasoning benchmarkClaude Opus 5.518.1 ppHLE: Claude 61.4% vs DeepSeek 43.3%12
Output speed (tokens/s)DeepSeek V4.1 Flash 101772.2 tokens/sDeepSeek 169.0 vs Claude 96.8 tokens/s12
Estimated request costDeepSeek V4.1 Flash 1017$0.0558DeepSeek $0.0042 vs Claude $0.060034

Coding and reasoning benchmarks

MetricDeepSeek V4.1 Flash 1017 (Max)Claude Opus 5.5 (Max)
AA · Terminal-Bench 4.01233.8%59.6%
AA · HLE1243.3%61.4%
AA · SciCode1255.9%66.9%

Performance

MetricDeepSeek V4.1 Flash 1017 (Max)Claude Opus 5.5 (Max)
Output speed (tokens/s)12169.096.8
Time to first answer token (s)1213.0676.8

Prices and request cost

MetricDeepSeek V4.1 Flash 1017Claude Opus 5.5
Input / 1M tokens34$0.30$4.00
Output / 1M tokens34$1.20$20.00
Context window (tokens)341,048,5761,000,000
Cost example: 10,000 input + 1,000 billed output tokens
DeepSeek V4.1 Flash 1017$0.0042
Claude Opus 5.5$0.0600

When each option fits

DeepSeek V4.1 Flash 1017

  • For high-volume pipelines, Novita AI lists DeepSeek V4.1 Flash 1017 at $0.30 input and $1.20 output per 1M tokens, making the example request about 14× cheaper than Claude.34
  • For streaming-heavy interfaces, DeepSeek’s AA output speed of 169.0 tokens/s is 1.7× Claude’s 96.8 tokens/s, and its median time to first answer token is 13.0 s versus 676.8 s.12
  • Its largest gap is agentic coding: 33.8% versus 59.6% on Terminal-Bench 4.0 (25.8 pp), compared with 11.0 pp on SciCode.12

Claude Opus 5.5

  • For terminal-based coding agents, Claude’s 59.6% on Terminal-Bench 4.0 is the clearest reason to pay its $4.00/$20.00 per-1M-token rates.123
  • For reasoning-heavy tasks, Claude reaches 61.4% on HLE versus DeepSeek’s 43.3%, an 18.1 pp lead.12
  • Context does not separate them: Novita AI lists 1,048,576 tokens for DeepSeek V4.1 Flash 1017 and Anthropic lists 1,000,000 for Claude.34

Methodology and sources

AA uses the named reasoning configurations; Claude Opus 5.5 is Max with default fallback.123

Coding and reasoning rows use the same AA evaluation versions: Terminal-Bench 4.0, HLE and SciCode. The available official reports do not provide this exact current model pair under a common setup. AA output speed and time to first answer token are model references, not Novita endpoint guarantees; the latter is AA’s median for its 10,000-input-token prompt and includes reasoning for both models. Prices and context windows are Novita AI’s listed values for DeepSeek V4.1 Flash 1017 and Anthropic’s for Claude Opus 5.5; the uncached example uses 10,000 input and 1,000 billed output tokens.

Costs are estimates at listed rates for uncached synchronous requests, not measured invoices; billed output includes reasoning. Tax, retries, discounts and negotiated capacity are excluded. Novita AI publishes this page using public sources, without a private workload benchmark.

  1. Artificial Analysis · DeepSeek V4.1 Flash 1017 (Max)
  2. Artificial Analysis · Claude Opus 5.5 (Max, Default Fallback)
  3. Anthropic · Opus 5.5 pricing
  4. Novita AI · DeepSeek V4.1 Flash 1017 pricing

Build with Novita Model APIs

Create an API key and connect your application.

Start BuildingExplore Models