neomanex
D

DeepSeek V4 Pro

DeepSeek · Released Jul 2026

View pricingCompare models
Recommended

Strong price-performance on open MIT weights; the 2x peak window is 01:00-04:00 and 06:00-10:00 UTC on weekdays only, so weekends bill entirely off-peak.

Is it right for you?

Good for

  • Cost-sensitive coding and autonomous agent workloads: far cheaper per session than GPT-5.5 or Opus 4.8 at competitive quality, even after the August 2026 price rise
  • Self-hosted enterprise and regulated-industry deployment: MIT license means proprietary code never leaves your infrastructure
  • Competitive programming and code-generation benchmarks: 93.5% LiveCodeBench, 3206 Codeforces Elo
  • Weekend and overnight batch work: peak billing applies only Monday to Friday, so all weekend hours bill at the off-peak $0.66 / $1.98 rate

Not good for

  • Terminal and CLI agent tasks: trails GPT-5.5 (67.9% vs 75.1% on Terminal-Bench 2.0)
  • Workloads requiring Western data-residency guarantees without self-hosting: DeepSeek's API is China-hosted
  • Weekday workloads pinned to peak hours: rates double to $1.32 / $3.96 during 01:00-04:00 and 06:00-10:00 UTC, Monday through Friday only

Pricing

Input

$0.66 / 1M tokens

Output

$1.98 / 1M tokens

Context

1M tokens. Peak 2x: 01-04+06-10 UTC Mon-Fri

View full pricing

Benchmarks

BenchmarkScoreSource
SWE-bench Verified80.6% Source
LiveCodeBench (Pass@1)93.5% Source
Codeforces Elo3206 Source
Terminal-Bench 2.067.9% Source
SWE Pro (Resolved)55.4% Source
MCPAtlas Public73.6% Source
GPQA Diamond71.7 Source

Verdict history

Sep 9, 2026
recommended -> recommended — No rating change. Rates re-verified unchanged against the vendor rate card on 9 September 2026. The peak window is Monday to Friday, not daily, so our "seven daily peak hours" summary overstated the premium window by 14 hours a week. The standing summary is corrected; the rating stays recommended.
Aug 20, 2026
recommended -> recommended (pricing premise corrected, benchmark error fixed) — Flagged for re-review on 2026-08-15 when DeepSeek announced the hike. Vendor pricing confirmed 2026-08-20: off-peak $0.66/$1.98, peak $1.32/$3.96, effective 2026-08-16, replacing the flat $0.435/$0.87. The "3x cheaper than Fable 5 per completed task" claim still holds with margin but was dropped rather than restated, since the Fable 5 rate was not re-verified in this pass. Separately, the old summary claimed 99.2% on SWE-bench Verified, contradicting this entity's own benchmark row of 80.6%; corrected to 80.6%. The unsourced "Claude Code parity on CursorBench 3.2" claim was also removed. Rating held at recommended.

Sources

Verification log

  • Pricing— Updated

    Automated agent

    Rates re-verified UNCHANGED: off-peak $0.66 in / $1.98 out, peak $1.32 / $3.96 per 1M; cache hits $0.022 off-peak / $0.044 peak. Peak window re-confirmed verbatim as 01:00-04:00 and 06:00-10:00 UTC, Monday through Friday, matching the correction this entity already received on 2026-09-07; no further prose change needed and no verdict payload staged (its 2026-09-09 transition already landed). ONE defect fixed: pricing_url was stored without a trailing slash and resolves to 'Your First API Call | DeepSeek API Docs', a quickstart carrying no prices; the rate card is the same path with the slash. Corrected to https://api-docs.deepseek.com/quick_start/pricing/ so the citation supports the figures it is attached to. Rating unchanged at recommended. TOOL DEFECT OBSERVED: the sources array passed on model_update was resolved idempotently by URL against an existing Source row, so the supplied label and access_date 2026-09-09 were discarded and the row still reads label 'DeepSeek API Docs - Models & Pricing (V4 Pro: $0.44/$0.87)' with access_date 2026-07-18. That label states a pre-August rate that is no longer live.

  • Pricing— Updated

    Automated agent

    Per-token rates re-verified and UNCHANGED: off-peak $0.66 in / $1.98 out, peak $1.32 / $3.96 per 1M. The peak-window SCOPE differs from what we stored: the vendor page states 'Peak hours are 01:00 - 04:00 and 06:00 - 10:00 UTC, Monday through Friday (all other hours are off-peak).' Our pricing_context and buying guidance omitted the weekday qualifier and the verdict summary described 'seven daily peak hours', overstating the premium window by 14 hours a week (35 not 49). Refreshed pricing_context to '1M tokens. Peak 2x: 01-04+06-10 UTC Mon-Fri', corrected the peak strength to name Monday-Friday and the doubled rates, added a weekend/off-peak best_for row and a matching for_whom audience. The 2026-08-20 pass already removed the false '3x cheaper than Fable 5' multiple, so no false price multiple remains in the summary. Rating unchanged at recommended; verdict summary wording fix staged for the announcing draft 'deepseek-v4-peak-billing-weekdays-only'.

  • Pricing— No changes

    Automated agent

  • Pricing— No changes

    Automated agent

How we evaluate