Across 20 comparable benchmarks, DeepSeek-V4.1-Flash has better reported results on 11 and worse on 9. 2 further shared results are listed below, each with the reason it is not counted.
Moderate evidence20 of 22 shared results countedHow we compare
DeepSeek-V4.1-Flash is no longer available through its official API. API prices for this version are not shown.
Price basis: International rates vs listed rates. The price section below shows the applicable periods.
| Fact | MiMo-V2.6-Flash | DeepSeek-V4.1-Flash |
|---|---|---|
| Context window | 1M tokens | 1M tokens |
| Input / 1M | $0.140 | β |
| Output / 1M | $0.280 | β |
| Blended (4:1) | $0.069 | β |
| Status | Available | Retired |
| Released | September 22, 2026 | September 10, 2026 |
Blended is 4:1 input to output with 90% cache hits at listed rates β set your own mix.
Editorβs take
LLMLearner Editorial Β· Figures as of Β· AI-assisted
This is Xiaomi MiMo V2.6 Flash, not MiMo Pro. It offers a useful budget baseline, but the DeepSeek side is the September checkpoint rather than the model currently served by the Flash API.
Which Flash is being compared? The Xiaomi side is MiMo V2.6 Flash. The DeepSeek side is the original V4.1 Flash, retired from the official API after the October 17 upgrade. Historical weights and published evaluations can remain useful even when an API alias changes. For the larger Xiaomi model, use DeepSeek V4.1 Flash vs MiMo V2.6 Pro.
Use specific tasks to read the results. Xiaomi reports 67.9 on DeepSWE v1.1 and 52.3 on AutomationBench for Flash. The September DeepSeek release reports 74.2 and 54.8 respectively. These are provider reports, not our controlled trial. The shared-result table adds sources and evaluation settings; do not substitute an OSWorld score for evidence about repository repair. Try the same small set of failing repository tasks and multi-step workflows with equal tool permissions and output budgets.
Make the price assumptions explicit. The catalog's MiMo Flash international quote is $0.14 input/$0.28 output per million tokens, checked October 31. With 100,000 uncached input tokens and 10,000 output tokens, that is $0.0168 before tools and retries. The current DeepSeek Flash API costs $0.042 at peak or $0.021 off-peak for the same volume. These DeepSeek rates buy 1017; they are not an offer to rent the retired September model. Confirm the MiMo quote for your service region, then use the price controls below to model your own cache hits and output mix.
Choose a budget baseline, then test failure cost. MiMo Flash merits a trial when many inexpensive, checkable tasks dominate your workload. A model that needs extra attempts can lose its rate advantage. If Pro is also on your shortlist, the MiMo Flash vs MiMo Pro comparison isolates that upgrade decision. For today's DeepSeek behavior, start from the 1017 model record and select the exact counterpart.
Mean of 16 shared benchmarks grouped by capability on a 0β100 scale. How averages are built
DeepSeek-V4.1-Flash ahead on 4 of 6 counted
MiMo-V2.6-Flash ahead on 1 of 1 counted
DeepSeek-V4.1-Flash ahead on 2 of 3 counted
Split 3β3 of 6 counted
DeepSeek-V4.1-Flash ahead on 2 of 3 counted
No shared comparable results for Abstract reasoning, Computer & tool use, Research & search, Long context, Multimodal.
Published scores, with matching modes where available. Gaps show MiMo-V2.6-Flash relative to DeepSeek-V4.1-Flash. Reading the results
| Benchmark | MiMo-V2.6-Flash | DeepSeek-V4.1-Flash | Gap |
|---|---|---|---|
| Code MigrationMaintenance & Optimization Β· Hidden-test pass rate (%) | 40.9Unspecified Reasoning Β· tools | 45.6β Thinking High Β· tools | β4.7 pts#31 vs #18 of 79 |
| DeepSWERepository Engineering Β· Pass@1 (DeepSWE v1.1) | 67.9Thinking Β· tools | 74.2β Max Β· tools | β6.3 pts#31 vs #11 of 73 |
| IOI (Vals v2)Algorithmic Coding Β· Subtask points (%) | 47.7Unspecified Reasoning Β· tools | 40.3β Thinking High Β· tools | +7.4 pts#28 vs #35 of 45 |
| ProgramBenchCode Generation & Editing Β· Almost resolved (>=95%) | 26Thinking Β· toolsΒ· programbench-almost-resolved | 20.3β Max Β· toolsΒ· programbench-almost-resolved | +5.7 pts#14 vs #17 of 23 |
| Terminal-Bench 2.1Agentic Development Β· Accuracy | 87.6Thinking Β· tools | 90.6β Max Β· tools | β3.0 pts#16 vs #5 of 68 |
| Terminal-Bench 4.0Agentic Development Β· Resolution rate (%)Not counted Β· Different or unrecorded test environments β excluded from the tally | 28.8Thinking Β· toolsΒ· legacy-default | 35.4β Max Β· toolsΒ· vals-effort-study | β#39 vs #33 of 90 |
| Vibe Code Bench v1.1Code Generation & Editing Β· Pass rate (%) | 79.0Unspecified Reasoning Β· tools | 84.7β Thinking High Β· tools | β5.8 pts#34 vs #23 of 121 |
| ProofBench v1.1 (Lean 4)Mathematics Β· Accuracy (%) | 63.0Unspecified Reasoning Β· tools | 54.0β Thinking High Β· tools | +9.0 pts#24 vs #31 of 54 |
| MysteryMechanismScientific Reasoning Β· Accuracy (%) | 21.6Unspecified Reasoning Β· tools | 21.2β Thinking High Β· tools | +0.4 pts#25 vs #26 of 32 |
| MedCodeClinical Workflows Β· Accuracy (%) | 41.1Unspecified Reasoning Β· tools | 41.2β Thinking High Β· tools | β0.1 pts#55 vs #54 of 97 |
| MedScribeClinical Workflows Β· Rubric pass rate (%) | 85.3Unspecified Reasoning Β· tools | 85.5β Thinking High Β· tools | β0.2 pts#29 vs #26 of 99 |
| AutomationBenchOffice & Business Β· Pass RateNot counted Β· The test environment is not recorded β excluded from the tally | 52.3Thinking Β· tools | 54.8β Max Β· tools | β#8 vs #5 of 64 |
| EMB (Excel Modeling)Finance Β· Accuracy (%) | 65.5Unspecified Reasoning Β· tools | 57.2β Thinking High Β· tools | +8.2 pts#26 vs #42 of 78 |
| Finance Agent v2Finance Β· Score | 56.3Unspecified Reasoning Β· tools | 53.5β Thinking High Β· tools | 1.05Γ#20 vs #33 of 75 |
| Legal Research BenchLegal Β· Score | 38.0Unspecified Reasoning Β· tools | 41.4β Thinking High Β· tools | 0.92Γ#44 vs #32 of 91 |
| Public Benefits Bench v1.1Service Workflows Β· Criterion pass rate (%) | 67.6Unspecified Reasoning Β· tools | 64.3β Thinking High Β· tools | +3.3 pts#14 vs #23 of 50 |
| SAGEService Workflows Β· Accuracy (%) | 43.5Unspecified Reasoning Β· tools | 47.9β Thinking High Β· tools | β4.4 pts#48 vs #29 of 82 |
| Tax Agent BenchFinance Β· Accuracy (%, rubric Γ citation quality) | 28.5Unspecified Reasoning Β· tools | 62.5β Thinking High Β· tools | β34.0 pts#43 vs #15 of 74 |
| CyberGymVulnerability Analysis Β· Score | 95.1Thinking Β· tools | 88.1β Max Β· tools | 1.08Γ#1 vs #3 of 14 |
| ExploitGymExploitation Β· Score (%) | 6.0Thinking Β· toolsΒ· exploitgym-budget-unspecified | 15.3β Max Β· toolsΒ· exploitgym-budget-unspecified | β9.3 pts#10 vs #8 of 12 |
| SEC-Bench ProVulnerability Analysis Β· Score (%) | 47.5Thinking Β· tools | 62.8β Max Β· tools | β15.3 pts#12 vs #10 of 13 |
| Vals Index v2.1Capability Indices Β· GDP-weighted accuracy (%) | 53.2Reported best | 51.3β Thinking High Β· tools | +1.9 pts#17 vs #23 of 38 |
Maintenance & Optimization Β· Hidden-test pass rate (%)
Repository Engineering Β· Pass@1 (DeepSWE v1.1)
Algorithmic Coding Β· Subtask points (%)
Code Generation & Editing Β· Almost resolved (>=95%)
Agentic Development Β· Accuracy
Agentic Development Β· Resolution rate (%)
Not counted Β· Different or unrecorded test environments β excluded from the tally
Code Generation & Editing Β· Pass rate (%)
Mathematics Β· Accuracy (%)
Scientific Reasoning Β· Accuracy (%)
Clinical Workflows Β· Accuracy (%)
Clinical Workflows Β· Rubric pass rate (%)
Office & Business Β· Pass Rate
Not counted Β· The test environment is not recorded β excluded from the tally
Finance Β· Accuracy (%)
Finance Β· Score
Legal Β· Score
Service Workflows Β· Criterion pass rate (%)
Service Workflows Β· Accuracy (%)
Finance Β· Accuracy (%, rubric Γ citation quality)
Vulnerability Analysis Β· Score
Exploitation Β· Score (%)
Vulnerability Analysis Β· Score (%)
Capability Indices Β· GDP-weighted accuracy (%)
A dash means no result is available.
Time-of-day providers use peak-hour rates; fixed-price providers use standard rates. Off-peak discounts remain in the published details.
Drag the handle above the bar for cache hits, the one below it for the input : output mix Β· 5% steps
Blended price = effective input rate at 90% cache hits Γ 80% + output rate Γ 20%
A blended cost needs a published input and output price on both sides, and one of these two does not have them yet.
MiMo-V2.6-Flash: International pricing. Xiaomi MiMo international pricing β input Xiaomi MiMo international pricing β output
DeepSeek-V4.1-Flash: Listed text rates; pricing tier not confirmed.
| Price / 1M tokens | MiMo-V2.6-Flash | DeepSeek-V4.1-Flash | Ratio |
|---|---|---|---|
| Input | $0.14 | β | β |
| Output | $0.28 | β | β |
| Cached input | $0.003 | β | β |
| Cache write | β | β | β |
| Blended | $0.069 | β | β |
Estimates use the selected token mix. Cache writes, tool fees and long-context surcharges are excluded.
Reported cost per task, paired with the score from the same run. Settings and sources may differ; costs do not affect the verdict.
MiMo-V2.6-Flash relative to DeepSeek-V4.1-Flash
Summary: 1 of 3 shared benchmarks have matching recorded settings and sources. The average is a geometric mean of 1 positive-cost ratios, one equal weight per benchmark. This benchmark set can include related tests and indexes; it is not a workload cost estimate.
MiMo-V2.6-Flash 2.4Γ cheaper
Settings or sources differ
Settings or sources differ
MiMo-V2.6-Flash 2.4Γ cheaper
Settings or sources differ
Settings or sources differ
Settings or sources differ or are not recorded; interpret the comparison with care.
MiMo-V2.6-Flash
DeepSeek-V4.1-Flash
Effort, protocol, aggregation and source match; other run conditions may differ.
MiMo-V2.6-Flash
DeepSeek-V4.1-Flash
Settings or sources differ or are not recorded; interpret the comparison with care.
MiMo-V2.6-Flash
DeepSeek-V4.1-Flash
| Benchmark | MiMo-V2.6-Flash | DeepSeek-V4.1-Flash |
|---|---|---|
| AA Intelligence Index v5.0.1 | Not reported | |
| AA-LCR | Not reported | |
| Agent Arena | Not reported | $0.1479 / task Median Β· Net improvement index (percentage points vs. average model): 2.3 Max (With Tools + Internet) |
| ARC-AGI-1 | Not reported | 1 more settings |
| ARC-AGI-2 | Not reported | 1 more settings |
| Code Migration | $0.4936 / task Average Β· Hidden-test pass rate (%): 40.9 Unspecified Reasoning (With Tools) | Not reported |
| CWE-bench v1 | Not reported | |
| EMB (Excel Modeling) | Not reported | |
| Finance Agent v2 | Not reported | |
| FrontierCode 2.0 | Not reported | $0.274 / task Average Β· Weighted rubric score: 18.2 Thinking High (With Tools + Internet) $0.206 / task Average Β· Weighted rubric score: 16.0 Thinking Low (With Tools + Internet) |
| FrontierCS 2 | Not reported | |
| IOI (Vals v2) | Not reported | |
| Legal Research Bench | Not reported | |
| LiveBench 2.0 | Not reported | $0.1669 / task Average Β· Overall score (equal-weight mean of 7 suite scores, 0-100): 78.0 Max (task-specific tools) |
| MedCode | Not reported | |
| MedScribe | $0.002184 / task Average Β· Rubric pass rate (%): 85.3 Unspecified Reasoning (With Tools) | Not reported |
| MysteryMechanism | Not reported | |
| ProofBench v1.1 (Lean 4) | Not reported | |
| Public Benefits Bench v1.1 | $0.0318 / task Average Β· Criterion pass rate (%): 67.6 Unspecified Reasoning (With Tools) | Not reported |
| SAGE | Not reported | |
| Terminal-Bench 4.1 | Not reported | |
| Vals Index v2.2 | $0.2216 / task Average Β· GDP-weighted accuracy (%): 54.8 Reported best (effort unspecified) | Not reported |
| Vibe Code Bench 1-100 | Not reported | |
| Vibe Code Bench v1.1 | Not reported | |
| WeirdML v3 | Not reported | $0.7371 / task Average Β· Official score (0-1; higher is better): 0.059 Thinking High (With Tools) |
| Attribute | MiMo-V2.6-Flash | DeepSeek-V4.1-Flash |
|---|---|---|
| Availability | Available | Retired |
| Provider | Xiaomi | DeepSeek-AI |
| Model type | Multimodal model | Multimodal model |
| First release | September 22, 2026 | September 10, 2026 |
| Context window | 1M tokens | 1M tokens |
| Max output | β | 384K tokens |
| Total parameters | 309B | 552B |
| Active parameters | 15B | 16B |
| Architecture | Mixture of Experts (MoE) | Mixture of Experts (MoE) |
| Documented capabilities | Reasoning, Multilingual, Coding, Tool use, Function calling, Computer use, Long context | Reasoning, Multilingual, Coding, Function calling, Tool use, Structured output, Long context |
| Input modalities | Text, Image, Audio, Video | Text, Image |
| Output modalities | Text | Text |
| Weights | Open weights | Open weights |
| Licence | MIT License | MIT License |
| Commercial use | Commercial use allowed | Commercial use allowed |
These are reported-score comparisons, not controlled head-to-head tests.
What people found comparing these models, or using either one. Comments also appear on the models' own pages.
Used MiMo-V2.6-Flash or DeepSeek-V4.1-Flash? Sign in with Google to share what worked, what did not, and which API or setup you tested.
No community comments on either model yet.
Alternatives with the most benchmark overlap for this matchup.