以下部分引用 AI 总结，现阶段 AI 仍然有幻觉。请以内容的原文为准。
数据新鲜度群聊 07-14 ✓卖方 07-14 ✓Wrap 断档15天 ✕News 07-15 ✓X 断档12天 ✕页面生成 07-15
Xiaomi's MiMo V2.5 Pro
22 atoms · 跨 1 天 · 首见 2026-04-24 · 最近 2026-04-24
三色: 🟦 fact 19 · 🟥 take 3 · stance ▲11/▼1/◆0
来源: X 22
时态: fresh:22
标签: 好数字:17 · 好观点:3 · 好信源:2
预测 (forecast) (1)
- 2026-04-24 · weight release claim
weights are expected to be released soon, which would make MiMo V2.5 Pro the first equal open weights model - slightly ahead of DeepSeek V4 Pro
tags: 好观点 · → daily
事实 (fact) (21)
- 2026-04-24 · Intelligence Index score
MiMo V2.5 Pro has landed at 54 in the Artificial Analysis Intelligence Index
tags: 好数字 · → daily
value: qty=54 · unit=score
- 2026-04-24 · weights release status
MiMo V2.5 Pro’s weights are expected to be released soon
tags: 好信源 · → daily
value: date=soon
- 2026-04-24 · Improved score from MiMo V2 Pro
MiMo V2.5 Pro shows an impressive improvement over MiMo V2 Pro (49)
tags: 好数字 · → daily
value: qty=5 · unit=points
- 2026-04-24 · Release date
Release date: April 22, 2026
tags: 好数字 · → daily
value: date=2026-04-22
- 2026-04-24 · Parameters
Parameters: 1T total, 42B active
tags: 好数字 · → daily
value: qty=1T total, 42B active · unit=parameters
- 2026-04-24 · Context window
Context window: 1M tokens
tags: 好数字 · → daily
value: qty=1M · unit=tokens
- 2026-04-24 · API pricing
Priced at $1.00/$3.00 per million input/output tokens on Xiaomi’s first-party API
tags: 好数字 · → daily
value: qty=$1.00/$3.00 per million input/output tokens · unit=USD
- 2026-04-24 · Cost efficiency claim
It costs only $462 to run the Artificial Analysis Intelligence Index, compared to $948 for Kimi K2.6 and $544 for GLM 5.1
tags: 好数字 · → daily
value: qty=$462 · unit=USD
- 2026-04-24 · AA-Omniscience Index score
MiMo V2.5 Pro scores 4 on the AA-Omniscience Index
tags: 好数字 · → daily
value: qty=4 · unit=score
- 2026-04-24 · Hallucination rate
MiMo V2.5 Pro demonstrates a relatively low hallucination rate (25%)
tags: 好数字 · → daily
value: qty=25% · unit=rate
- 2026-04-24 · Accuracy
low accuracy (23%)
tags: 好数字 · → daily
value: qty=23% · unit=rate
- 2026-04-24 · Agentic benchmark score
scores 1578 on GDPval-AA, ahead of DeepSeek V4 Pro (1554), GLM-5.1 (1535), MiniMax-M2.7 (1514), and Kimi K2.6 (1484)
tags: 好数字 · → daily
value: qty=1578 · unit=score
- 2026-04-24 · HLE score improvement
The model scores 34% on HLE (+6% from MiMo V2.0)
tags: 好数字 · → daily
value: qty=34% · unit=rate
- 2026-04-24 · IFBench score improvement
80% on IFBench (+11% from MiMo V2.0)
tags: 好数字 · → daily
value: qty=80% · unit=rate
- 2026-04-24 · CritPt regression
a small regression in CritPt (5% to 4%)
tags: 好数字 · → daily
value: qty=5% to 4% · unit=rate
- 2026-04-24 · Token efficiency vs peers
MiMo V2.5 Pro's token efficiency remains competitive against peers in a similar intelligence tier, using ~92M output tokens for the Intelligence Index
tags: 好数字 · → daily
value: qty=~92M · unit=tokens
- 2026-04-24 · Cost-efficiency vs competitors
It costs only $462 to run the Artificial Analysis Intelligence Index, compared to $948 for Kimi K2.6 and $544 for GLM 5.1
tags: 好数字 · → daily
value: qty=$462 · unit=USD
- 2026-04-24 · Agentic benchmark score claim
The model scores 1578 on GDPval-AA, and places it in the top tier for real-world work tasks among recent releases
tags: 好数字 · → daily
value: qty=1578 · unit=score
- 2026-04-24 · Evaluation breakdown available
Breakdown of individual evaluation results for MiMo V2.5 Pro
tags: 好信源 · → daily
- 2026-04-24 · Token efficiency claim
MiMo V2.5 Pro is considerably more token efficient than models in the same Intelligence tier
tags: 好观点 · → daily
- 2026-04-24 · Agentic leadership claim
MiMo V2.5 Pro leads its peer group on agentic tasks
tags: 好观点 · → daily
时间轴 (近 20)
- 2026-04-24 ·
fact · Intelligence Index score · → daily
- 2026-04-24 ·
fact · weights release status · → daily
- 2026-04-24 ·
forecast · weight release claim · → daily
- 2026-04-24 ·
fact · Improved score from MiMo V2 Pro · → daily
- 2026-04-24 ·
fact · Release date · → daily
- 2026-04-24 ·
fact · Parameters · → daily
- 2026-04-24 ·
fact · Context window · → daily
- 2026-04-24 ·
fact · API pricing · → daily
- 2026-04-24 ·
fact · Cost efficiency claim · → daily
- 2026-04-24 ·
fact · AA-Omniscience Index score · → daily
- 2026-04-24 ·
fact · Hallucination rate · → daily
- 2026-04-24 ·
fact · Accuracy · → daily
- 2026-04-24 ·
fact · Agentic benchmark score · → daily
- 2026-04-24 ·
fact · HLE score improvement · → daily
- 2026-04-24 ·
fact · IFBench score improvement · → daily
- 2026-04-24 ·
fact · CritPt regression · → daily
- 2026-04-24 ·
fact · Token efficiency vs peers · → daily
- 2026-04-24 ·
fact · Cost-efficiency vs competitors · → daily
- 2026-04-24 ·
fact · Token efficiency claim · → daily
- 2026-04-24 ·
fact · Agentic leadership claim · → daily
← 实体目录 · 系统日志