Скорость вывода
28154.1 tok/s
30149.7 tok/s
34Qwen3140.3 tok/s
35133.4 tok/s
Qwen3 is available through 237 API providers on LMSpeed. Compare API pricing from $0.0010 to $75.00 per million input tokens across providers. Free API access is offered by 6 providers. Context window: 262,144. In speed benchmarks, the fastest provider reaches 1214 tok/s.
Alibaba Qwen3 is the Qwen family's flagship LLM series with dense and MoE variants, seamless thinking/non-thinking modes, and leading open-source performance in math, code, and agent tasks.
Также известна как
Compare Qwen3 API pricing across 231 providers. Input prices range from $0.0010 to $75.00 per million input. 10dian-API offers the lowest rate at $0.0010/M. 6 providers offer free API credits or a free tier.
| Провайдер | Вариант модели | Аудит | Входные данные ($/M) | Выходные данные ($/M) | Скорость (т/с) | Первый токен |
|---|---|---|---|---|---|---|
ModelPool97.7% | qwen3-8b | — | $0.0061-96% | $0.024-94% | 76.0 t/s | 0.98 s-12% |
| qwen3-4b | — | $0.0037-98% | $0.015-96% | — | — | |
| qwen3-14b | — | $0.012-92% | $0.049-88% | — | — | |
| qwen3-32b | — | $0.024-84% | $0.098-76% | — | — | |
V-API100% | qwen3-235b-a22b-c | — | $0.040-73% | $0.040-90% | — | — |
| qwen3-4b | — | $0.300 | $1.20 | — | — | |
| qwen3-8b | — | $0.500 | $2.00 | — | — | |
| qwen3-30b-a3b | — | $0.750 | $3.00 | — | — | |
| qwen3-14b | — | $1.00 | $4.00 | — | — | |
| qwen3-235b-a22b | — | $2.00 | $8.00 | — | — | |
| qwen3-32b | — | $2.00 | $8.00 | — | — | |
老张API100% | qwen3-30b-a3b | — | $0.192 | $1.92 | — | — |
| qwen3-32b | — | $0.384 | $3.84 | — | — | |
| qwen3-235b-a22b | — | $0.959 | $9.59 | — | — | |
柏拉图AI100% | qwen3-4b | — | $0.041-73% | $0.164-59% | — | — |
| qwen3-8b | — | $0.068-54% | $0.274-32% | — | — | |
| qwen3-30b-a3b | — | $0.104-31% | $0.411 | — | — | |
| qwen3-14b | — | $0.137-9% | $0.548 | — | — | |
| qwen3-235b-a22b | — | $0.274 | $1.10 | — | — | |
| qwen3-32b | — | $0.274 | $1.10 | — | — |
Данные о ценах из публичных API провайдеров
Сравните производительность по скорости и задержке у всех API-провайдеров.
| Провайдер | Скорость | Задержка | Тесты |
|---|---|---|---|
RinkoAI Qwen/Qwen3-32B | 1214.07 tok/s | 0.91s | 5 |
qwen-3-32b | 1028.76 tok/s | 0.97s | 20 |
qwen-3-235b | 686.99 tok/s | 1.62s | 5 |
qwen-3-235b-2507 | 625.49 tok/s | 0.37s | 5 |
Qwen/Qwen3-235B | 234.72 tok/s | 1.71s | 25 |
qwen/qwen3-32b | 143.07 tok/s | 13.67s | 15 |
qwen3:0.6b | 122.09 tok/s | 0.73s | 15 |
Qwen/Qwen3-235B-A22B | 116.62 tok/s | 21.27s | 30 |
qwen3-30b-a3b | 114.09 tok/s | 0.45s | 5 |
qwen3-8b | 103.87 tok/s | 7.64s | 15 |
qwen/qwen3-30b-a3b | 88.00 tok/s | 13.53s | 60 |
qwen/qwen3-30b-a3b:free | 83.19 tok/s | 22.71s | 5 |
unsloth/qwen3:30b-a3b-q8_0 | 82.71 tok/s | 2.21s | 5 |
qwen3-8b | 78.65 tok/s | 10.01s | 5 |
qwen3-8b | 76.04 tok/s | 0.98s | 5 |
qwen3-14b | 75.24 tok/s | 9.23s | 5 |
/root/models/Qwen/Qwen3-4B | 72.15 tok/s | 0.56s | 5 |
qwen/qwen3-14b | 69.57 tok/s | 18.89s | 5 |
Qwen/Qwen3-30B-A3B | 64.95 tok/s | 18.06s | 45 |
qwen3 | 64.49 tok/s | 17.71s | 5 |
Показано 1–20 из 42 провайдеров
deepseek-v3-2
DeepSeek V3.2 is an upgraded V3-series MoE model with stronger reasoning, coding, and math performance, widely available through OpenAI-compatible API relays.
kimi-k2-5
Moonshot Kimi K2.5 is an open-weight multimodal agent model with native vision and text input, strong coding performance, and a 256K context window.
minimax-m2-5
MiniMax M2.5 is MiniMax's flagship text model for coding and agents, with SOTA-level programming and agentic performance, improved token efficiency, and fast high-TPS API deployment.
glm-5
Zhipu GLM-5 is Zhipu flagship GLM series model with enhanced reasoning, agent capabilities, and strong performance on Chinese enterprise and coding scenarios.
gpt-5-4
OpenAI GPT-5.4 extends the GPT-5 family with stronger instruction following, deeper tool use, and improved performance on coding, math, and long-document analysis.
glm-4-7
Zhipu GLM-4.7 is a flagship GLM release from Zhipu AI with advanced Chinese-English reasoning, coding, and agent features.
Рейтинги основаны на тестах, предоставленных сообществом, и периодических зондах работоспособности. Носит рекомендательный характер, не является официальными данными.Artificial Analysis
