MiniMaxLanguage model
MiniMax-M2.5-highspeed
MiniMax-M2.5-highspeed has no overall index value yet because there are too few independent benchmark results. From the vendor-reported results we estimate about 75.8 points (95% range 56.5 to 95.0). This is not an independent measurement.
PopularityRank 678 of 785
22out of 100+9 vs. last month
based on two data groups, mostly search
Summary
MiniMax-M2.5-highspeed names the faster API serving path for MiniMax M2.5. MiniMax describes both versions as producing identical results, with Highspeed adding speed. The label provides no separate context or output specification.
The label describes only the serving layer for response time and throughput. MiniMax does not establish a separate checkpoint through this API name.
All benchmark results
Every published result with its source, test setup and comparison models.
Show all 14 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 Verified (Show measurement, test conditions, and source)
- Source value
- 80.2
- Score
- 80.2
- Metric
- resolved rate as reported
- Unit
- %
- Benchmark version
- SWE-Bench Verified; Claude Code; overridden default system prompt; four-run average
- Category
- coding
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- MiniMax M2.5 official release benchmark results
- Status
- active
- Retrieved at
- 2026-10-15
- Methodology
- Vendor-reported result. Exact M2.5 publisher results; official MiniMax model documentation states M2.5 and M2.5-highspeed have identical quality results.
- Context
- M2.5-highspeed is not separately evaluated here: MiniMax explicitly bridges that API alias to identical M2.5 results. SWE tests use Claude Code with overridden default prompt and four-run average; harness details differ by task.
SWE-Bench Pro (Show measurement, test conditions, and source)
- Source value
- 55.4
- Score
- 55.4
- Metric
- score as reported
- Unit
- %
- Benchmark version
- SWE-Bench Pro; Claude Code; overridden default system prompt; four-run average
- Category
- coding
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- MiniMax M2.5 official release benchmark results
- Status
- active
- Retrieved at
- 2026-10-15
- Methodology
- Vendor-reported result. Exact M2.5 publisher results; official MiniMax model documentation states M2.5 and M2.5-highspeed have identical quality results.
- Context
- M2.5-highspeed is not separately evaluated here: MiniMax explicitly bridges that API alias to identical M2.5 results. SWE tests use Claude Code with overridden default prompt and four-run average; harness details differ by task.
SWE-Bench Multilingual (Show measurement, test conditions, and source)
- Source value
- 74.1
- Score
- 74.1
- Metric
- score as reported
- Unit
- %
- Benchmark version
- SWE-Bench Multilingual; Claude Code; overridden default system prompt; four-run average
- Category
- coding
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- MiniMax M2.5 official release benchmark results
- Status
- active
- Retrieved at
- 2026-10-15
- Methodology
- Vendor-reported result. Exact M2.5 publisher results; official MiniMax model documentation states M2.5 and M2.5-highspeed have identical quality results.
- Context
- M2.5-highspeed is not separately evaluated here: MiniMax explicitly bridges that API alias to identical M2.5 results. SWE tests use Claude Code with overridden default prompt and four-run average; harness details differ by task.
Multi-SWE-Bench (Show measurement, test conditions, and source)
- Source value
- 51.3
- Score
- 51.3
- Metric
- score as reported
- Unit
- %
- Benchmark version
- Multi-SWE-Bench; Claude Code; overridden default system prompt; four-run average
- Category
- coding
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- MiniMax M2.5 official release benchmark results
- Status
- active
- Retrieved at
- 2026-10-15
- Methodology
- Vendor-reported result. Exact M2.5 publisher results; official MiniMax model documentation states M2.5 and M2.5-highspeed have identical quality results.
- Context
- M2.5-highspeed is not separately evaluated here: MiniMax explicitly bridges that API alias to identical M2.5 results. SWE tests use Claude Code with overridden default prompt and four-run average; harness details differ by task.
Terminal Bench 2 (Show measurement, test conditions, and source)
- Source value
- 51.7
- Score
- 51.7
- Metric
- success score as reported
- Unit
- %
- Benchmark version
- Terminal Bench 2; Claude Code 2.0.64; 8-core/16GB; timeout 7200s; four-run average
- Category
- coding
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- MiniMax M2.5 official release benchmark results
- Status
- active
- Retrieved at
- 2026-10-15
- Methodology
- Vendor-reported result. Exact base M2.5 publisher result; MiniMax states M2.5-highspeed has identical quality. Terminal Bench 2 uses Claude Code 2.0.64, 8-core/16GB sandbox and 7200-second timeout, average of four runs.
- Context
- The official API model page states M2.5-highspeed has identical quality results to M2.5; this remains a publisher claim, not a highspeed-specific rerun.
BrowseComp (Show measurement, test conditions, and source)
- Source value
- 76.3
- Score
- 76.3
- Metric
- score with context management
- Unit
- %
- Benchmark version
- BrowseComp; MiniMax WebExplorer-like agent; context history discarded above 30% maximum context
- Category
- agentic
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- MiniMax M2.5 official release benchmark results
- Status
- active
- Retrieved at
- 2026-10-15
- Methodology
- Vendor-reported result. Exact base M2.5 BrowseComp result with context management; MiniMax states M2.5-highspeed has identical quality. The release does not provide a pinned task snapshot.
- Context
- MiniMax's official model list states identical M2.5 and M2.5-highspeed quality; no independent alias evaluation is implied.
CodeAgentBench (Show measurement, test conditions, and source)
- Source value
- 30.5
- Score
- 30.5
- 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-M2.5-highspeed; page updated 2026-06-18; 151 tasks, three attempts per task (453 attempts total; source reports 46/151 solved at least once, 22/151 in all three attempts, 102/453 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
- 14.6
- Score
- 14.6
- 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-M2.5-highspeed; page updated 2026-06-18; 151 tasks, three attempts per task (453 attempts total; source reports 46/151 solved at least once, 22/151 in all three attempts, 102/453 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.
MiniMax Provider Verifier: Query-Success-Rate (Show measurement, test conditions, and source)
- Source value
- 100
- Score
- 100
- Metric
- Query-Success-Rate: probability of eventually returning a valid response within max_retry=10, as defined by the source.
- Unit
- %
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- MiniMax-AI MiniMax-Provider-Verifier
- Status
- active
- Retrieved at
- 2026-10-21
- Methodology
- Vendor-reported result. The source labels this as February 2026 data for the openRouter-minimax-highspeed provider route. It says each initial-release test prompt was executed 10 times per provider and reports the mean over those runs. The prompt count per metric and an immutable serving revision are not stated.
- Context
- Vendor-reported measurements of a named third-party provider deployment, not a general result for every MiniMax-M2.5 serving path. The ten-run repetition is not a per-metric sample count; no confidence interval or immutable serving revision is published.
MiniMax Provider Verifier: ToolCalls-Trigger Similarity (Show measurement, test conditions, and source)
- Source value
- 99.22
- Score
- 99.22
- Metric
- ToolCalls-Trigger Similarity: F1 similarity of third-party tool-call triggering against the official MiniMax deployment.
- Unit
- %
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- MiniMax-AI MiniMax-Provider-Verifier
- Status
- active
- Retrieved at
- 2026-10-21
- Methodology
- Vendor-reported result. The source labels this as February 2026 data for the openRouter-minimax-highspeed provider route. It says each initial-release test prompt was executed 10 times per provider and reports the mean over those runs. The prompt count per metric and an immutable serving revision are not stated.
- Context
- Vendor-reported measurements of a named third-party provider deployment, not a general result for every MiniMax-M2.5 serving path. The ten-run repetition is not a per-metric sample count; no confidence interval or immutable serving revision is published.
MiniMax Provider Verifier: Language-Following-Success-Rate (Show measurement, test conditions, and source)
- Source value
- 80
- Score
- 80
- Metric
- Language-Following-Success-Rate in minor-language scenarios; the source describes it as a rate of valid language following.
- Unit
- %
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- MiniMax-AI MiniMax-Provider-Verifier
- Status
- active
- Retrieved at
- 2026-10-21
- Methodology
- Vendor-reported result. The source labels this as February 2026 data for the openRouter-minimax-highspeed provider route. It says each initial-release test prompt was executed 10 times per provider and reports the mean over those runs. The prompt count per metric and an immutable serving revision are not stated.
- Context
- Vendor-reported measurements of a named third-party provider deployment, not a general result for every MiniMax-M2.5 serving path. The ten-run repetition is not a per-metric sample count; no confidence interval or immutable serving revision is published. The source notes that this metric is sensitive to top-k and related decoding parameters. The February 2026 table does not disclose those settings.
GDPval-MM (Show measurement, test conditions, and source)
- Source value
- 59
- Score
- 59
- Metric
- average win rate
- Unit
- %
- Direction
- No direction published
- Source type
- vendor-reported
- Evaluator
- MiniMax M2.5 release announcement
- Status
- active
- Retrieved at
- 2026-10-24
- Methodology
- Vendor-reported result. MiniMax M2.5 result bridged to the M2.5-highspeed API version by MiniMax's statement that the two API versions have identical results; internal Cowork Agent evaluation based on the open-source GDPval test set, with an LLM judge making pairwise win/tie/loss judgments on complete trajectories.
- Context
- The release page does not report an evaluation sample count or confidence interval. Attribution to M2.5-highspeed relies solely on the vendor's API-version equivalence statement; no separate highspeed evaluation is reported.
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-M2.5-highspeedMiniMax M2.7 22Fugu Max 22o3 22Voxtral Small 24B 22
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-M2.5-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-M2.5-highspeed
How good is MiniMax-M2.5-highspeed in benchmarks?
Independent results are not yet enough for the overall index. An estimate from 3 vendor-reported results (SWE-bench Pro (public) 55.4%, SWE-bench Verified 80.2%, Terminal-Bench 2.0 51.7%) gives about 75.8 points, with a 95% range of 56.5 to 95.0. The estimate is calibrated on models with both vendor-reported and independent results. Vendor numbers run clearly above independent measurements on some tests; the calibration offsets that on average.
Vendor estimate, not an overall index value
How large is the context window of MiniMax-M2.5-highspeed?
204,800 tokens. That is how much text MiniMax-M2.5-highspeed reads at once, about 310 A4 pages.
Page count as a rule of thumb from Artificial Analysis
MiniMax-M2.5-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-M2.5-highspeed locally?
Yes, the weights are open. With 229B parameters it needs a lot of memory, more a server than a laptop.
Stated by MiniMax
- Open weights
- yes
- License
- Modified MIT (MiniMax)
- Parameters
- 229B
- active per token
- n/a
Sources and data date
Show all 14 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.