Decision support

DeepSeek V4 Flash Vision Exp vs Hy4

Compare canonical objects directly. The useful dimensions change with the object type, so model and hardware comparisons stay separate.

Models Hardware

DeepSeek V4 Flash Vision Exp

DeepSeek

Sources:models.devOpenRouter

Hy4

Tencent

Sources:models.devOpenRouter

Benchmark results

Agent Arena
IPS score · Higher is better
- DeepSeek V4 Flash Vision Exp
  Not reported
- Hy4
  +0.053
  Arena #14CI +0.040 to +0.066
Runs and sources
DeepSeek V4 Flash Vision Exp
Not reported
Hy4
default+0.05395% CI +0.040 to +0.066 · 3,304,477 observations
Source:Agent Arena leaderboard, 2026-11-11

Each model's best published result is shown; "Not reported" means no matched evaluation, not a score of zero. Arena and DeepSWE ranks are the leaderboards' own; other ranks are among catalogue models with a score. Results from different harnesses or settings are not directly comparable.

Arena ratings: Arena (lmarena-ai/leaderboard-dataset), licensed CC BY 4.0. Ratings come from pairwise human votes and are only comparable within one leaderboard; the interval is Arena's 95% confidence interval. Model names are matched to the Vertical Tabs catalog.

Model specifications

Release date

August 21, 2026

November 4, 2026

Context window

Native context

—

Max output tokens

943,717

65,536

Knowledge cutoff

—

Open weights

Parameter count

305B

—

API pricing · USD per 1M tokens

Input

$0.44

$0.417

Output

$1.32

$1.2505

Cached input

$0.014

$0.021

Cache write

—

Capabilities and modalities

Input modalities

text, image

Output modalities

text

Tool calling

Supported

Reasoning option

Supported

Supported options

frequency_penalty, include_reasoning, logit_bias, logprobs, max_tokens, min_p, presence_penalty, reasoning, reasoning_effort, repetition_penalty, response_format, seed, stop, structured_outputs, temperature, tool_choice, tools, top_k, top_logprobs, top_p

include_reasoning, max_completion_tokens, max_tokens, reasoning, reasoning_effort, response_format, stop, structured_outputs, temperature, tool_choice, tools

Capabilities

—

Architecture and availability

Architecture

—

Model families

—

Tokenizer

DeepSeek

Other

Instruction format

—

License

MIT

—

Providers

Provider endpoints

Ollama variants

Metadata source

ai-research

A dash means the field is not reported. Provider-specific performance and pricing may differ.

Benchmark results

Agent Arena

Model specifications

API pricing · USD per 1M tokens

Capabilities and modalities

Architecture and availability