OpenCode 榜单 — AI 编程模型用量与成本排名
OpenCode 榜单
用量时序、成本、缓存、份额、地区、留存 · 对齐源站全部板块
用量第一
DeepSeek V4 Flash
25.6T tokens
最便宜
MIMO V2.5
$0.00/次
涨最快
Muse Spark 1.2 Contributor
+27%
份额第一
DeepSeek
92.4%
Top Models · 时序
Leaderboard · 模型用量榜
· 点击表头排序#
模型
1
DeepSeek V4 Flash
DeepSeek
25.6T
$0.03
+24%
94.2%
2
Muse Spark 1.2 Contributor
Meta
13.5T
$0.01
+27%
91.4%
3
MIMO V2.5
Xiaomi
9.7T
$0.00
-23%
93%
4
GLM 5.3 Flash
Zhipu
3.8T
$0.32
-92%
94.3%
5
Nemotron 3 Ultra
NVIDIA
3.0T
—
-28%
83.2%
6
DeepSeek V4 Flash Vision Exp
DeepSeek
2.0T
$0.25
-15%
98%
7
Muse Spark 1.3 Contributor
Meta
1.7T
$0.03
%
90%
8
Hy3
Tencent
1.6T
$0.01
-54%
91.8%
9
DeepSeek V4 Pro
DeepSeek
1.3T
$0.27
-40%
96.3%
10
Nemotron 3.5 Lightning
NVIDIA
710B
—
-17%
81.8%
11
GPT 5.6 Luna
OpenAI
534B
—
-36%
—
12
Qwen3.8 Flash
Qwen
489B
—
%
—
13
Ling 3.0 Flash Fin
Unknown
345B
—
%
—
14
Minimax M3
MiniMax
326B
—
-26%
—
15
GLM 5.2
Zhipu
203B
—
-42%
—
16
Laguna S 2.1
Unknown
189B
—
-25%
—
17
MIMO V2.5 Pro
Xiaomi
188B
—
-23%
—
18
GLM 5.3
Zhipu
165B
—
-52%
—
Session Cost · 会话成本
· 单次任务中位数成本，最省在前| # | 模型 | $/会话 | Tokens |
|---|---|---|---|
| 1 | MIMO V2.5 | $0.00 | 858.1T |
| 2 | Hy3 | $0.01 | 978.2T |
| 3 | Muse Spark 1.2 Contributor | $0.01 | 4349.8T |
| 4 | Muse Spark 1.3 Contributor | $0.03 | 7402.9T |
| 5 | DeepSeek V4 Flash | $0.03 | 5473.7T |
| 6 | DeepSeek V4 Flash Vision Exp | $0.25 | 18176.1T |
| 7 | DeepSeek V4 Pro | $0.27 | 10622.8T |
| 8 | GLM 5.3 Flash | $0.32 | 8245.2T |
Token Cost · Token 价格
· 每百万 token 单价（$/M），最省在前| # | 模型 | 加权 | 输入 | 输出 | 缓存 |
|---|---|---|---|---|---|
| 1 | Nemotron 3.5 Lightning | $0.200 | $0.080 | $0.200 | $0.040 |
| 2 | GLM 5.3 Flash | $0.250 | $0.075 | $0.250 | $0.015 |
| 3 | DeepSeek V4 Flash | $0.280 | $0.140 | $0.280 | $0.028 |
| 4 | MIMO V2.5 | $0.280 | $0.140 | $0.280 | $0.003 |
| 5 | Hy3 | $0.580 | $0.140 | $0.580 | $0.035 |
| 6 | DeepSeek V4 Flash Vision Exp | $1.320 | $0.440 | $1.320 | $0.028 |
| 7 | DeepSeek V4 Pro | $3.200 | $1.600 | $3.200 | $0.135 |
Cache Ratio · 缓存命中率
· 命中越高成本越省| # | 模型 | 缓存率 | 命中 / 未命中 |
|---|---|---|---|
| 1 | DeepSeek V4 Flash Vision Exp | 98% | 2.0k / 40.1 |
| 2 | DeepSeek V4 Pro | 96.3% | 1.2k / 46.5 |
| 3 | GLM 5.3 Flash | 94.3% | 3.5k / 213 |
| 4 | DeepSeek V4 Flash | 94.2% | 23.9k / 1.5k |
| 5 | MIMO V2.5 | 93% | 9.0k / 675.6 |
| 6 | Hy3 | 91.8% | 1.5k / 131.4 |
| 7 | Muse Spark 1.2 Contributor | 91.4% | 12.3k / 1.2k |
| 8 | Muse Spark 1.3 Contributor | 90% | 1.5k / 164.9 |
| 9 | Nemotron 3 Ultra | 83.2% | 2.5k / 510.6 |
| 10 | Nemotron 3.5 Lightning | 81.8% | 575.7 / 127.9 |
Retention · 用户留存榜
· 模型周留存率（回访 / 可回访）| # | 模型 | 留存率 | 样本周 |
|---|---|---|---|
| 1 | DeepSeek V4 Flash Vision Exp DeepSeek | 89.3% | 1.2k |
| 2 | Muse Spark 1.2 Contributor Meta | 85.7% | 1.8k |
| 3 | GLM 5.3 Flash Zhipu | 84.5% | 5.0k |
| 4 | Qwen3.5 Plus Qwen | 73.5% | 268 |
| 5 | Kimi K2.5 Moonshot | 72.1% | 875 |
| 6 | MIMO V2.5 Xiaomi | 71.9% | 18.8k |
| 7 | DeepSeek V4 Flash DeepSeek | 69.2% | 418.9k |
| 8 | Hy3 Tencent | 69.1% | 1.5k |
| 9 | Minimax M2.7 MiniMax | 68.8% | 3.6k |
| 10 | GLM 5.3 Zhipu | 67.2% | 7.4k |
| 11 | Kimi K2.6 Moonshot | 65.7% | 10.0k |
| 12 | Minimax M2.5 MiniMax | 65.3% | 340 |
| 13 | DeepSeek V4 Pro DeepSeek | 65.2% | 136.8k |
| 14 | Minimax M3 MiniMax | 65.2% | 27.8k |
| 15 | MIMO V2.5 Pro Xiaomi | 65.2% | 8.1k |