Claude Haiku 5.5 vs DeepSeek V4 Pro (0813).
Verdict by category, tale of the tape, shared benchmarks only, and cost at three workloads — all from the tracked data.
Verdict by category
| Category | Claude Haiku 5.5 | DeepSeek V4 Pro (0813) | Winner |
|---|---|---|---|
| League Score | 66.2 | 63.2 | Claude Haiku 5.5 |
| Coding (SWE-bench %) | — | 73.6 | — |
| Reasoning (GPQA %) | — | 90.1 | — |
| Context (tokens) | 1000000 | 1000000 | Draw |
| Speed (tok/s) | — | 99.4 | — |
| Price (blended $/1M) | $0.20 | $1.98 | Claude Haiku 5.5 (cheaper) |
Tale of the tape
| Spec | Claude Haiku 5.5 | DeepSeek V4 Pro (0813) |
|---|---|---|
| Maker | Anthropic | DeepSeek |
| Released | Oct 7, 2026 | Apr 24, 2026 |
| League Score | 66 | 63 |
| Context | 1,000,000 tokens | 1,000,000 tokens |
| Max output | 128,000 | 384,000 |
| $/1M input | $0.10 | $1.32 |
| $/1M output | $0.50 | $3.96 |
| Weights | Closed | Open |
| Speed | — | 99.4 t/s |
Head-to-head benchmarks (shared only)
| Benchmark | Claude Haiku 5.5 | DeepSeek V4 Pro (0813) |
|---|---|---|
| No shared benchmark measurements yet — shown as not measured, never interpolated. | ||
Cost at three workloads
| Workload | Claude Haiku 5.5 | DeepSeek V4 Pro (0813) |
|---|---|---|
| Light — 1M in / 250K out | $0.23 | $2.31 |
| Standard — 10M in / 2.5M out | $2.25 | $23.10 |
| Long-context — 50M in / 5M out | $7.50 | $85.80 |
The pick
Pick Claude Haiku 5.5 if you want the higher overall League Score; you want the lower running cost.
Pick DeepSeek V4 Pro (0813) if you want downloadable weights / self-hosting; coding is the priority.
All figures from the model database — sources and dates on each model page. Cost your own workload or compare a different pair.
Frequently asked questions
Which is better overall, Claude Haiku 5.5 or DeepSeek V4 Pro (0813)?
Claude Haiku 5.5 leads on overall League Score, 66 to 63 — a gap of 3 points on a composite of sourced benchmarks. That is overall form, not every category: check the per-category verdicts before choosing.
Which is cheaper, Claude Haiku 5.5 or DeepSeek V4 Pro (0813)?
Claude Haiku 5.5 is cheaper: $0.20 vs $1.98 blended per 1M tokens (3:1 input/output mix) — about 9.9x less per token.
Which has the bigger context window, Claude Haiku 5.5 or DeepSeek V4 Pro (0813)?
Neither — both take 1M of context.
Are Claude Haiku 5.5 and DeepSeek V4 Pro (0813) open source?
Claude Haiku 5.5 is closed weights — API access only; DeepSeek V4 Pro (0813) is open weights — it can be downloaded and self-hosted.