head-to-head · model vs model
GPT-4o vs DeepSeek V4 Flash
Live matchup data for GPT-4o and DeepSeek V4 Flash: OpenRouter usage, pricing, benchmark scores, and community sentiment, refreshed daily. No verdicts — read the tape.
share of openrouter tokens
as of Jul 7, 2026 · Δ vs ≥7d prior
usage share — openrouter model rankings
Usage crossover
source: OpenRouter model rankings · daily · 90d · no OpenRouter feed: GPT-4o
through Jul 7, 2026
side by side
Spec & pricing
| Spec | GPT-4o | DeepSeek V4 Flash |
|---|---|---|
| Vendor | OpenAI | DeepSeek |
| Homepage | openrouter.ai | openrouter.ai |
| Context window | 128K tokens | 1M tokens |
| Input price | $0.15/MTok | $0.09/MTok |
| Output price | $0.6/MTok | $0.18/MTok |
| Cache read | — | $0.018/MTok |
| Price as of | Jul 8, 2026 · openrouter | Jul 8, 2026 · openrouter |
sources: provider pages · OpenRouter · GitHub · package registries
— = not tracked / not applicable
head to head
Benchmarks
| Benchmark | GPT-4o | DeepSeek V4 Flash | Source |
|---|---|---|---|
| aa-agentic-index | 1 index— · openai/gpt-4o-mini | 31.1 index— · deepseek/deepseek-v4-flash | openrouter.ai |
| aa-coding-index | 24.2 index— · openai/gpt-4o-2024-05-13 | 56.2 index— · deepseek/deepseek-v4-flash | openrouter.ai |
| aa-intelligence-index | — | 40.3 index— · deepseek/deepseek-v4-flash | openrouter.ai |
| aider-edit | 72.9%May 13 · gpt-4o-2024-05-13 | — | github.com |
| aider-o1-polyglot | 15.1%Dec 21 · gpt-4o-2024-11-20 | — | github.com |
| Aider Polyglot | 23.1%Dec 30 · gpt-4o-2024-08-06 | — | github.com |
| aider-refactor | 62.9%May 13 · gpt-4o | — | github.com |
| SWE-bench Lite | 38%Jun 27 · abanteai mentatbot gpt4o | — | github.com |
| swe-bench-multimodal | 30.8%May 31 · GUIRepair gpt4o | — | github.com |
| swe-bench-test | 18.8%Jun 28 · autocoderover v20240620 | — | github.com |
| SWE-bench Verified | 27%Oct 16 · epam ai run gpt 4o | — | github.com |
best tracked score per benchmark · scores are not directly comparable across benchmarks
sources linked per row
community pulse
Reddit sentiment
| Signal | GPT-4o | DeepSeek V4 Flash |
|---|---|---|
sentiment feed pending — Reddit pulse lands with the daily crawl | ||
source: Reddit mentions, LLM-classified sentiment · trailing 30d
awaiting first crawl
faq
Frequently asked
- Is GPT-4o more used than DeepSeek V4 Flash?
- The Token Tape tracks daily OpenRouter usage for this matchup, but GPT-4o has no usage feed yet, so no usage comparison is published. Pricing, benchmarks, and sentiment compare on the available signals.
- How does GPT-4o pricing compare to DeepSeek V4 Flash?
- GPT-4o runs $0.15 per million input tokens and $0.6 per million output tokens (as of Jul 8, 2026). DeepSeek V4 Flash runs $0.09 per million input tokens and $0.18 per million output tokens (as of Jul 8, 2026). Full plan and per-token detail is in the spec table above, each with its source and capture date.
- How do GPT-4o and DeepSeek V4 Flash score on coding benchmarks?
- On aa-agentic-index, GPT-4o scores 1 index (—) and DeepSeek V4 Flash scores 31.1 index (—). All tracked results and their sources are in the benchmark table above.
- Where does this comparison data come from?
- OpenRouter public rankings (usage), provider pricing pages (plans and per-token prices), GitHub and package registries (adoption), public benchmark leaderboards, and Reddit (sentiment). Every stored datum keeps its source URL, and the tape refreshes daily. The Token Tape publishes the data without declaring a winner.
alerts
Get the weekly tape — usage moves, price changes, and new benchmark results for tracked matchups.