MiniMaxLanguage model
MiniMax-M2.7-highspeed
MiniMax-M2.7-highspeed has no overall index value yet because there are too few independent benchmark results.
PopularityRank 600 of 785
27out of 100-1 vs. last month
based on two data groups, mostly search
Summary
MiniMax-M2.7-highspeed is the faster API identifier for M2.7 serving. MiniMax describes M2.7 and M2.7-highspeed as two API versions with identical results but higher speed. MiniMax M2.7 supplies the documented capabilities, open weights, and agent focus.
Highspeed concerns throughput and latency within the provider platform. MiniMax’s statement about identical results does not establish whether both API versions use the same released checkpoint.
All benchmark results
Every published result with its source, test setup and comparison models.
Show all 13 published benchmark results
Page 1 of 2
Measurements without matching peer values
These measurements have no matching peer values under the same test conditions. Their original values and sources remain available here.
SWE-Bench Pro (Show measurement, test conditions, and source)
- Source value
- 56.22
- Score
- 56.22
- Metric
- score as reported
- Unit
- %
- Benchmark version
- SWE-Bench Pro; page does not state full harness configuration
- Category
- coding
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- MiniMax M2.7 official model benchmark page
- Status
- active
- Retrieved at
- 2026-10-15
- Methodology
- Vendor-reported result. Exact MiniMax-M2.7 model page values; official MiniMax model documentation states M2.7 and M2.7-highspeed have identical results.
- Context
- M2.7-highspeed has no separate published run; MiniMax's model page is the only score bridge for the alias. SWE-Pro and Terminal Bench settings are not fully given here.
Terminal Bench 2 (Show measurement, test conditions, and source)
- Source value
- 57
- Score
- 57
- Metric
- score as reported
- Unit
- %
- Benchmark version
- Terminal Bench 2; page does not state full harness configuration
- Category
- coding
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- MiniMax M2.7 official model benchmark page
- Status
- active
- Retrieved at
- 2026-10-15
- Methodology
- Vendor-reported result. Exact MiniMax-M2.7 model page values; official MiniMax model documentation states M2.7 and M2.7-highspeed have identical results.
- Context
- M2.7-highspeed has no separate published run; MiniMax's model page is the only score bridge for the alias. SWE-Pro and Terminal Bench settings are not fully given here.
VIBE-Pro (Show measurement, test conditions, and source)
- Source value
- 55.6
- Score
- 55.6
- Metric
- end-to-end project-delivery score as reported
- Unit
- %
- Benchmark version
- VIBE-Pro; MiniMax internal benchmark
- Category
- agentic
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- MiniMax M2.7 official model benchmark page
- Status
- active
- Retrieved at
- 2026-10-15
- Methodology
- Vendor-reported result. Exact publisher-defined VIBE-Pro score for M2.7; API docs state the M2.7-highspeed alias has identical results.
- Context
- VIBE-Pro is MiniMax's internal benchmark, not an independent public leaderboard. M2.7-highspeed has no separate published run; the alias mapping is a vendor assertion.
MiniMax complex skill adherence (Show measurement, test conditions, and source)
- Source value
- 97
- Score
- 97
- Metric
- Skill adherence rate
- Unit
- %
- Category
- agentic-capability
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- MiniMax official model page
- Status
- active
- Retrieved at
- 2026-10-16
- Methodology
- Vendor-reported result. MiniMax's current official M2.7 page reports a 97% skill-adherence rate on 40 complex-skill cases exceeding 2,000 tokens. It does not define whether the 40 cases are the score denominator or explain the aggregation, and does not publish a confidence interval. The page states M2.7-highspeed has identical results but reports no separate highspeed run.
- Context
- Vendor-reported narrow skill-adherence result, not a general model-quality score. The source's 40 cases are not assumed to equal the score denominator. Highspeed attribution relies on MiniMax's equivalence statement, not a separate highspeed run.
CodeAgentBench (Show measurement, test conditions, and source)
- Source value
- 25.2
- Score
- 25.2
- Metric
- Pass@3
- Unit
- %
- Category
- coding
- Direction
- No direction published
- Sample
- 151
- Source type
- community-reported
- Evaluator
- CodeAgentBench public community result page
- Status
- active
- Retrieved at
- 2026-10-19
- Methodology
- Community-reported result. opencode-cli 1.14.32; model route minimax-cn-coding-plan/MiniMax-M2.7-highspeed; page updated 2026-06-18; 151 tasks, three attempts per task (453 attempts total; source reports 38/151 solved at least once, 15/151 in all three attempts, 76/453 attempts). Temperature/sampling settings and immutable served revision are not stated.
- Context
- Public community-reported OpenCode run, not an independent evaluator report. The page does not state a temperature/sampling protocol or immutable served revision; its verifier audit rechecked only part of the initial successes.
CodeAgentBench (Show measurement, test conditions, and source)
- Source value
- 9.9
- Score
- 9.9
- Metric
- Pass^3
- Unit
- %
- Category
- coding
- Direction
- No direction published
- Sample
- 151
- Source type
- community-reported
- Evaluator
- CodeAgentBench public community result page
- Status
- active
- Retrieved at
- 2026-10-19
- Methodology
- Community-reported result. opencode-cli 1.14.32; model route minimax-cn-coding-plan/MiniMax-M2.7-highspeed; page updated 2026-06-18; 151 tasks, three attempts per task (453 attempts total; source reports 38/151 solved at least once, 15/151 in all three attempts, 76/453 attempts). Temperature/sampling settings and immutable served revision are not stated.
- Context
- Public community-reported OpenCode run, not an independent evaluator report. The page does not state a temperature/sampling protocol or immutable served revision; its verifier audit rechecked only part of the initial successes.
MLE Bench Lite (Show measurement, test conditions, and source)
- Source value
- 66.6
- Score
- 66.6
- Metric
- average medal rate
- Unit
- %
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- MiniMax official M2.7 benchmark report
- Status
- active
- Retrieved at
- 2026-10-20
- Methodology
- Vendor-reported result. MiniMax reports 22 MLE Bench Lite competitions across three 24-hour trials. Aggregation and task counts per competition are not stated. Attribution to M2.7-highspeed relies only on MiniMax's explicit identical-results statement.
- Context
- Vendor-reported for M2.7. MiniMax states that M2.7-highspeed has identical results, but this is not a separate highspeed run or independent reproduction.
NL2Repo (Show measurement, test conditions, and source)
- Source value
- 39.8
- Score
- 39.8
- Metric
- percent
- Unit
- %
- Direction
- No direction published
- Source type
- vendor-reported
- Evaluator
- MiniMax official M2.7 benchmark report
- Status
- active
- Retrieved at
- 2026-10-20
- Methodology
- Vendor-reported result. MiniMax reports 39.8% for M2.7 but does not name the precise metric, benchmark version, sample count, or run settings. Attribution to M2.7-highspeed relies only on MiniMax's explicit identical-results statement.
- Context
- Vendor-reported for M2.7. MiniMax states that M2.7-highspeed has identical results, but this is not a separate highspeed run or independent reproduction.
Toolathon (Show measurement, test conditions, and source)
- Source value
- 46.3
- Score
- 46.3
- Metric
- accuracy
- Unit
- %
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- MiniMax official M2.7 benchmark report
- Status
- active
- Retrieved at
- 2026-10-20
- Methodology
- Vendor-reported result. MiniMax reports 46.3% accuracy on Toolathon for M2.7. Benchmark version, sample count, and evaluation settings are not stated. Attribution to M2.7-highspeed relies only on MiniMax's explicit identical-results statement.
- Context
- Vendor-reported for M2.7. MiniMax states that M2.7-highspeed has identical results, but this is not a separate highspeed run or independent reproduction.
Terminal-Bench 2.0, MiniMax M2.7 Highspeed / Harness Agent (Show measurement, test conditions, and source)
- Source value
- 42.9213483146
- Score
- 42.9213483146
- Metric
- Resolution rate; the source API field is accuracy.
- Unit
- %
- Benchmark version
- Terminal-Bench 2.0; board 2-0; dataset 6d639009-20bb-47e4-87f7-56ef4d5fbc7f
- Category
- coding
- Direction
- Higher is better
- Confidence interval
- 2.9381642209
- Source type
- official-leaderboard
- Evaluator
- Terminal-Bench
- Status
- active
- Retrieved at
- 2026-10-28
- Methodology
- Official benchmark leaderboard. Official Terminal-Bench 2.0 leaderboard row; exact source model_names=["MiniMax-M2.7-highspeed"]; model label=MiniMax M2.7 Highspeed; model provider(s)=["minimax"]; model organization=MiniMax; agent=Harness Agent; agent version=unknown; agent organization=lazyFrogLOL; agent page/repository URL=https://github.com/lazyFrogLOL/Harness_Engineering; integration method=API; reasoning effort=not reported; board release-date field=2026-03-18; rank=77; status=display; row UUID=16d92deb-2fe5-4512-81d9-2dd3466cccef; board UUID=f58e955f-4db5-48a0-87ec-1c324a4a9414; dataset version=6d639009-20bb-47e4-87f7-56ef4d5fbc7f; source verified flag=false; n_trials field=0 sentinel; entry created_at=2026-08-28T21:43:32.982477+00:00; entry updated_at=2026-08-28T23:12:18.836648+00:00 (board-record timestamps, not trial start time).
- Context
- This is the published model-plus-agent system result, not an isolated model API score. Different agents, agent versions, integrations, or model endpoints are distinct configurations, not direct model-only comparisons. The leaderboard date is the row's listed release date, not a confirmed trial start date. The source returns n_trials=0 for these legacy rows; this is a sentinel, not zero measured trials, and no sample count is reported. The per-row verified boolean is preserved in configuration; false rows are still displayed by the official board, and this flag is not a confidence interval.
GDPval-AA (Show measurement, test conditions, and source)
- Source value
- 1495
- Score
- 1495
- Metric
- GDPval-AA Elo
- Unit
- elo
- Category
- knowledge-work
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- MiniMax official model page
- Status
- active
- Retrieved at
- 2026-10-16
- Methodology
- Vendor-reported result. MiniMax's current official M2.7 page reports 1495 Elo on GDPval-AA. It states that the M2.7 and M2.7-highspeed API versions have identical results, with highspeed faster. No separate highspeed evaluation, frozen revision, sample count, or confidence interval is published; attribution to the highspeed profile follows the vendor's equivalence claim only.
- Context
- Vendor-reported M2.7 result attributed to the highspeed API route only through MiniMax's statement that both versions have identical results. This is not a separate highspeed run or independent reproduction.
SWE Multilingual (Show measurement, test conditions, and source)
- Source value
- 76.5
- Score
- 76.5
- Metric
- unknown
- Unit
- No unit provided
- Direction
- No direction published
- Source type
- vendor-reported
- Evaluator
- MiniMax official M2.7 benchmark report
- Status
- active
- Retrieved at
- 2026-10-20
- Methodology
- Vendor-reported result. MiniMax reports this numeric value for M2.7 but does not state a named metric, unit, benchmark version, sample count, or run settings. Attribution to M2.7-highspeed relies only on MiniMax's explicit identical-results statement.
- Context
- Vendor-reported for M2.7. MiniMax states that M2.7-highspeed has identical results, but this is not a separate highspeed run or independent reproduction.
Page 1 of 2
Popularity compared
How often the model is searched for, used and mentioned, compared with every model of its category.
Popularity index
The 100 most popular models of this category, 0 to 100, higher is more popular
20 of 101 models
As of November 2026 · Source: gradually.ai popularity index
Over time
Weekly values from October 18, 2026 to November 7, 2026
MiniMax-M2.7-highspeedClaude Opus 4.7 27Aya Vision 8B 27Tiny Aya Fire 27Falcon 3 7B 26
Method changes in the timeline, a jump at these points is no market event:
- 1November 1, 2026: Value only with search or usage data, 30-day attention
What is gaining popularity right now, across models, vendors, harnesses and AI subscriptions, is collected in All AI trends.
Similar models
More models of the same kind as MiniMax-M2.7-highspeed, chosen by model line, developer, size, price and release date.
All specifications
Architecture
Input and output
Watermarking and provenance
AI watermarking refers to invisible signals embedded in the output. Visible marks and provenance metadata are listed separately. Unknown means there is insufficient evidence for that particular field.
- AI watermarking
- Unknown
- no source
- Watermarking technique
- Not known
- no source
Licence and access
Frequently asked questions about MiniMax-M2.7-highspeed
How large is the context window of MiniMax-M2.7-highspeed?
204,800 tokens. That is how much text MiniMax-M2.7-highspeed reads at once, about 310 A4 pages.
Page count as a rule of thumb from Artificial Analysis
MiniMax-M2.7-highspeedRank 79 of 103205K
MiniMax M2.7205K
MiniMax M2.5205K
MiniMax M2.1205K
GLM-4.7205K
GLM-4.6205K
GLM-5.1200K
Can I run MiniMax-M2.7-highspeed locally?
Yes, the weights are open. With 230B parameters it needs a lot of memory, more a server than a laptop.
Stated by MiniMax
- Open weights
- yes
- License
- MiniMax Non-Commercial License
- Parameters
- 230B
- active per token
- n/a
Sources and data date
Show all 15 sources
Benchmark sources at a glance
Which independent evaluators tested this model. This is coverage, not a quality ranking.
No active result from the reviewed external evaluators can currently be attributed to this model. Other source evidence remains available below.
Without vendor claims and historical or disputed results; those stay labelled in the single values.
| Type | Evidence and data date |
|---|---|
| Input | Documented modalities. Retrieved October 27, 2026. |
| Output | Documented modalities. Retrieved October 27, 2026. |
| Research status | Research date October 15, 2026. 6 source URLs checked. This documents the inspected sources, not an exhaustive inventory of every publication. |
| Not published | alias-specific-reproduction Note MiniMax asserts equivalence between M2.7 and M2.7-highspeed but publishes no separate highspeed-specific test. VIBE-Pro remains vendor-internal. |
| Not published | alias-specific-reproduction Note MiniMax states that M2.7 and M2.7-highspeed have identical results, but publishes no separate highspeed evaluation or independent reproduction. The observations here are assigned to this profile solely under that vendor claim. |
| Unresolved | benchmark-denominators-and-methodology Note The official M2.7 page does not expose the GDPval-AA sample count or setup. For skill adherence it names 40 long, complex skill cases but does not define the percentage denominator or aggregation. No sample count or confidence interval is inferred. |
| Additional source | Hugging Face · MiniMax M2.7 model card · Editorial description reviewed October 2, 2026 |
| Additional source | MiniMax M2.7 official model benchmark page (retrieved October 15, 2026; October 16, 2026; October 20, 2026) · MiniMax M2.7 model page · Editorial description reviewed October 2, 2026 |
| Additional source | CodeAgentBench public community result page (retrieved October 19, 2026) |
| Additional source | MiniMax official M2.7 benchmark report (retrieved October 20, 2026) |
| Additional source | Terminal-Bench (retrieved October 28, 2026) |
| Additional source | Model metadata (retrieved October 15, 2026) |
| Additional source | Model metadata (retrieved October 20, 2026) |
| Additional source | MiniMax API model list and pricing · Editorial description reviewed October 2, 2026 |
| Additional source | OpenRouter: Popularity index (derived), CC BY 4.0 |
| Additional source | Vercel AI Gateway: Popularity index (derived), CC BY 4.0 |
| Additional source | Ollama: cumulative downloads for the popularity index |
| Additional source | Replicate: cumulative runs for the popularity index |
| Additional source | Google Trending Now: limited feed selection for Germany and the US |
| Additional source | Creative Commons Attribution 4.0 |