Kimi K3 vs MiniMax M3

MiniMax M3 已从我们的目录下线。下方是它最后公布的价格;这个模型已无法调用,对比中的另一个模型仍可正常调用。

vs

什么时候选哪个

两者都接受文本、图像和视频输入并返回文本,双方都有对话、代码、推理和工具使用,所以真正的差别是价格和输出余量:kimi-k3 的输入贵 10 倍、输出贵 12.5 倍、缓存读取贵 5 倍。需要它 1048576 token 的上下文和同样 1048576 token 的最大输出,来一次性生成超长产物时选 kimi-k3,因为 minimax-m3 的输出上限是 524288。对成本敏感的批量、它的长上下文标志,以及关闭思考的选项(kimi-k3 不允许)则选 minimax-m3。

Benchmark 成绩

MiniMax M3:供应商没有公布 benchmark 成绩。

高于同类均值无更高分Kimi K352 / 6712 / 67
Kimi K3 MiniMax M3 其他参测模型 同类均值 ★ 没有模型得分更高
DeepSWE 1.1
67.5%
N/A
OSWorld-Verified
84.8%
N/A
Cybergym
80%
N/A
Finance Agent v2
54.4%
N/A
Harvey Lab-AA
没有模型得分更高 94.6%
N/A
GPQA Diamond
93.5%
N/A
BrowseComp
91.2%
N/A
Video-MME (w. sub)
没有模型得分更高 90%
N/A

供应商公布: Alibaba (Qwen) Alibaba (Qwen) Alibaba (Qwen) Alibaba (Qwen) Alibaba (Qwen) Alibaba (Qwen) Alibaba (Qwen) Anthropic Anthropic Anthropic Anthropic ByteDance DeepSeek DeepSeek Google Google Google Google MiniMax Moonshot Moonshot OpenAI OpenAI OpenAI OpenAI OpenAI Tencent Z.ai Z.ai Z.ai Z.ai

价格

Kimi K3 MiniMax M3 Δ
输入 / 1M token $3 $0.3 10×
输出 / 1M token $15 $1.2 13×
缓存读取 / 1M token $0.3 $0.06 5×
缓存写入 - 不单独收费 -

价格取自构建时的实时目录,最新价格见各模型页面。

两个模型所处的位置:全部 91 个按同一单位计费的聊天模型的每 1M token 输入价分布(对数刻度)

Kimi K3 · $3 MiniMax M3 · $0.3
$0.05 · Qwen3 VL Flash $30 · GPT-5.4 Pro

能力

Kimi K3 MiniMax M3
工具调用 是 是
思考控制 始终开启 可配置
结构化输出 是 -
提示词缓存 隐式(自动) 隐式(自动)
缓存有效期 未公布 未公布
最小缓存前缀 未公布 512 个 token

规格

Kimi K3 MiniMax M3
输入模态 文本 图像 视频 文本 图像 视频
输出模态 文本 文本
发布日期 2026-07-16 2026-06-01
上下文窗口 1,048,576 1,000,000
最大输出 1,048,576 524,288
思考参数 reasoning_effort (top-level; the thinking object is not accepted)
  • thinking.type
  • reasoning_split
可选值
reasoning_effort
  • low
  • high
  • max
thinking.type
  • adaptive
  • disabled
reasoning_split
  • boolean
默认值 max adaptive:思考开启,由模型判断何时需要额外推理

规格照录自各供应商的文档;供应商没有公布的项目,对应的行直接省略,不做推断。 完整来源: Kimi K3 · MiniMax M3

同一条提示词,两个模型,经网关实测

提示词 Explain to a CFO, in exactly three sentences, why our LLM bill doubled after we enabled reasoning. 检查 恰好 3 句

Kimi K3 通过 · 3 sentences

Reasoning models don't just answer questions—they "think" first, generating long internal chains of step-by-step logic before producing a response. Those hidden thinking steps are billed as output tokens (the most expensive kind, typically 3–5x the price of input tokens), and a single query can generate thousands of them even when the visible answer is only a paragraph long. So you're paying for dramatically more compute per request: the bill doubled because the model does far more work behind the scenes, not because usage increased.

输出 755 tok (+637 思考) 延迟 20.8 s

MiniMax M3 通过 · 3 sentences

Reasoning models don't just answer — they first generate extensive internal "thinking" tokens to work through the problem step-by-step, and you're billed for all of those tokens even though users never see them. A query that previously cost ~200 tokens may now cost 2,000–10,000 once reasoning is enabled, multiplying per-query spend by roughly 5–50x while the visible output stayed the same size — that's the doubling. The lever is tuning: cap reasoning effort on simple queries, cache the reasoning traces, and reserve the full reasoning model for only your hardest 10–20% of requests, which typically restores 60–80% of the savings.

输出 764 tok (+629 思考) 延迟 3.5 s

指令遵循(恰好三句,数得出来)、受众适配(对 CFO 说话的口吻),以及下方 token 计数暴露出的隐藏思考计费差额。

提示词 This function is supposed to return the longest run of consecutive equal items, but callers report it is off by one on single-element inputs and misses runs that end at the last index. Fix it and explain the bug in one paragraph. 检查 修复通过测试

Kimi K3 通过 · 8/8 cases

```python def longest_run(xs): if not xs: return 0 best = 1 cur = 1 for i in range(1, len(xs)): if xs[i] == xs[i - 1]: cur += 1 best = max(best, cur) else: cur = 1 return best ``` The bug is a fencepost error: `cur` counts *matching adjacent pairs* rather than *items in the run*, because it starts at 0 and only increments when `xs[i] == xs[i - 1]`. A run of length L contains L−1 equal pairs, so every run was undercounted by exactly one. For a single-element list (a run of length 1) the loop never runs and the function returns 0 instead of 1; likewise any run ending at the last index — e.g. the two `2`s in `[1, 2, 2]` — reported 1 instead of 2, making it look like no run existed at all. […]

输出 1837 tok (+1547 思考) 延迟 47.2 s

MiniMax M3 通过 · 8/8 cases

The bug is that the loop starts at index 1 and treats `cur` as the number of *equal transitions* rather than the run length, so a run of *k* consecutive equal items is reported as `k − 1`, and a single-element list (where the loop body never executes) returns 0 instead of 1. A correct version tracks the actual run length by seeding `cur = 1` and resetting it to 1 (not 0) on a mismatch: ```python def longest_run(xs): if not xs: return 0 best = cur = 1 for i in range(1, len(xs)): if xs[i] == xs[i - 1]: cur += 1 else: cur = 1 if cur > best: best = cur return best ``` Now `[5]` returns `1`, `[1,1,1]` returns `3`, and a run like `[1,2,2,2]` that ends at the last index is correctly captured as `3` because `cur` is updated before `best` is compared on every iteration, including the final one.

输出 845 tok (+596 思考) 延迟 3.5 s

修复是否真的正确(能运行)、解释的信息密度,以及在一个范围明确的任务上的 token 效率。

提示词 Extract a JSON object with fields {company, quarter, revenue_usd_m, growth_pct, guidance} from this text. Use null for anything not stated; add no other fields. 检查 合法 JSON,严格符合 schema

Kimi K3 通过 · 5/5 fields, guidance null

{ "company": "Meridian Robotics", "quarter": "September quarter", "revenue_usd_m": 412, "growth_pct": 11, "guidance": null }

输出 924 tok (+863 思考) 延迟 25.9 s

MiniMax M3 通过 · 5/5 fields, guidance null

```json { "company": null, "quarter": "Q3", "revenue_usd_m": 412, "growth_pct": null, "guidance": null } ```

输出 1002 tok (+955 思考) 延迟 3.6 s

是否严守 schema(不臆造字段)、能否顶住幻觉压力(原文明说暂不给出 guidance),以及结构化输出路径的差异。

提示词 Write a 120-word product announcement for an API gateway feature that caches prompts across providers. Forbidden words: "seamless", "unlock", "game-changer", "revolutionize", "empower". Exactly one sentence must be a question. 检查 120 词,0 个禁用词

Kimi K3 通过 · 120 words, 0 banned, 1 question

Meet Prompt Cache, a new API gateway capability that stores prompt responses and serves them across OpenAI, Anthropic, Google, and Azure endpoints. It matches requests by model, prompt hash, tools, temperature, and tenant policy, so repeated work returns fast while sensitive contexts stay isolated. Teams set TTLs, stale rules, encryption scopes, and bypass flags per route. Analytics show hit rate, latency saved, token spend avoided, and drift risk by provider. What changes for developers? Keep one integration, add cache headers, and watch fallback logic respect consent, residency, and audit needs. During rollout, canary keys compare fresh answers with cached copies before promotion. Prompt Cache cuts vendor calls, steadies p95 latency, and gives platform owners controls for cost, quality, and compliance.

输出 1527 tok (+1354 思考) 延迟 37.9 s

MiniMax M3 通过 · 121 words, 0 banned, 1 question

Introducing PromptCache, our intelligent caching layer for the API gateway. Tired of paying twice for identical requests? The feature stores prompt-response pairs across every connected provider, returning cached results instead of hitting costly endpoints. Deployment requires only a single configuration flag, and semantic similarity detection happens automatically. Latency drops up to 70% on repeated traffic while costs follow the same downward curve. Developers access detailed cache hit metrics, eviction logs, and provider breakdowns through the dashboard. Whether routing to OpenAI, Anthropic, or local models, every cached token saves time and money. The cache respects TTL policies, supports manual purging, and encrypts stored content. Available today for Pro and Enterprise tiers—enable PromptCache on existing endpoints and watch inference bills shrink immediately.

输出 2808 tok (+2656 思考) 延迟 6.0 s

是否守住约束(字数预算、禁用词表、只能有一个问句)、文风特征,以及长度控制。

改一行代码就能在两个模型之间切换

下方每个标签页里都有两个模型 ID,高亮的那两行是唯一要改的地方。端点不变,API key 不变,请求结构也不变。

from openai import OpenAI

client = OpenAI(
    base_url="https://synthorai.io/v1",
    api_key="sk-syn-...",
)

resp = client.chat.completions.create(
    model="kimi-k3",
    # model="minimax-m3",  # 取消注释此行,注释上一行
    messages=[{"role": "user", "content": "Summarize this diff"}],
    reasoning_effort="medium",
)
print(resp.choices[0].message.content)

获取 API key →

常见问题

Kimi K3 和 MiniMax M3 哪个更便宜?

按「输入 / 1M token」算,MiniMax M3 更便宜($0.3 对 $3,相差 10×)。其他计费项的结论可能相反,完整价格见上方表格,实际成本取决于你的用量构成。

不用分别集成两次,就能对 Kimi K3 和 MiniMax M3 做 A/B 测试吗?

可以。两个模型走同一个 OpenAI 兼容端点,用同一个 API key,切换时只要改一行里的模型名,所以可以给两个模型各分一部分流量,直接对比账单。

Kimi K3 和 MiniMax M3 支持提示词缓存吗?

支持。两个模型的缓存读取价都低于各自的输入价,所以前缀能反复命中缓存的负载,实际成本会比按官网价估算的低。具体的缓存读取价见上方价格表。

相关对比

我们的实测研究