Both are MiniMax models. MiniMax M3 is the newer, generally stronger default; reach for MiniMax M2.7 when a specific cost or latency profile matters more than the latest capabilities.
MiniMax M2.7 and MiniMax M3 are both MiniMax models, so the real question is not which lab to trust but which tier fits your workload and budget. MiniMax M2.7 is a cheap open-weight agentic coder with near-frontier terminal scores — held back by a non-commercial licence and non-standard benchmarks. MiniMax M3 is miniMax's open-weight 428B multimodal model with a 1M context — strong reported coding (59.0 SWE-Bench Pro) at low entry pricing. Since both come from the same lab, the comparison below focuses on the tier-and-cost trade-offs that actually separate them.
Key differences
Context window: MiniMax M3 holds 5.1× more — 1M (~1,573 pages) vs 205K (~307 pages). But effective recall usually fades long before the advertised ceiling, so the bigger number only helps if the model reasons over it.
Recency: MiniMax M3 is the newer model by about 3 months (released June 2026), usually meaning fresher training data and capabilities.
Specifications
Spec
MiniMax M2.7
MiniMax M3
Provider
MiniMax (China)
MiniMax (China)
Released
March 18, 2026
June 2026
Context window
205K (~307 pages)
1M (~1,573 pages)
Price (in/out)
$0.3/$1.2 per 1M tokens
$0.3/$1.2 per 1M tokens
Open weight?
Yes — self-hostable
Yes — self-hostable
Modalities
text, code
text, image, video, code
SWE-Bench Verified
Not published
Not published
MRCR v2 @ 1M
Not published
Not published
Who wins what
Agentic and terminal coding well above its price tier (57.0 on Terminal-Bench 2, vendor-reported): MiniMax M2.7 — MiniMax M3 is comparatively weak here — price doubles to $0.60/$2.40 above 512K tokens — not flat across 1M
Independently ranked 14th of 97 on the Artificial Analysis Intelligence Index: MiniMax M2.7 — MiniMax M3 is comparatively weak here — miniMax's own reported 80.5 SWE-Bench Verified figure is vendor-stated, not independently verified
Sparse mixture-of-experts — roughly 230B total but only ~10B active, so it runs on local hardware: MiniMax M2.7 — MiniMax M2.7 lists sparse mixture-of-experts — roughly 230B total but only ~10B active, so it runs on local hardware among its strengths; MiniMax M3 does not.
Open-weight 428B MoE (~23B active per token) with a 1M-token context: MiniMax M3 — Its 1M window holds about 5.1× more than MiniMax M2.7's 205K in a single prompt.
Native multimodal input — text, image and video: MiniMax M3 — MiniMax's open-weight 428B multimodal model with a 1M context — strong reported coding (59.0 SWE-Bench Pro) at low entry pricing — and it carries the larger 1M context.
Reports 59.0 on SWE-Bench Pro (a strong open-weight score), surpassing GPT-5.5: MiniMax M3 — MiniMax M2.7 is comparatively weak here — open weights but a NON-COMMERCIAL licence — commercial use requires prior written authorisation from MiniMax, and at least one major tracker still mislabels it as MIT
Largest single-prompt input: MiniMax M3 — Its 1M window is about 5.1× larger than MiniMax M2.7's 205K, fitting roughly 1,573 pages in one prompt.
Which should you pick?
Someone analysing very long documents or codebases: MiniMax M3 — Larger 1M window fits more in one prompt.
Anyone whose priority is agentic and terminal coding well above its price tier (57.0 on terminal-bench 2, vendor-reported): MiniMax M2.7 — It is specifically built for that.
Anyone whose priority is open-weight 428b moe (~23b active per token) with a 1m-token context: MiniMax M3 — That is its strongest area.
MiniMax M2.7: where it fits
A cheap open-weight agentic coder with near-frontier terminal scores — held back by a non-commercial licence and non-standard benchmarks. Released March 18, 2026 by MiniMax, it is built for agentic and terminal coding well above its price tier (57.0 on Terminal-Bench 2, vendor-reported), independently ranked 14th of 97 on the Artificial Analysis Intelligence Index, sparse mixture-of-experts — roughly 230B total but only ~10B active, so it runs on local hardware, and served by five separate hosts at uniform pricing, so there is no provider lock-in.
Its trade-offs are real: open weights but a NON-COMMERCIAL licence — commercial use requires prior written authorisation from MiniMax, and at least one major tracker still mislabels it as MIT, reports SWE-Bench Pro instead of the standard Verified set, which blocks like-for-like comparison, and already superseded internally by M3, and its 205K context is small against 1M-class rivals. At $0.3 in / $1.2 out per million tokens, it sits in the budget price band.
MiniMax M3: where it fits
MiniMax's open-weight 428B multimodal model with a 1M context — strong reported coding (59.0 SWE-Bench Pro) at low entry pricing. Released June 2026 by MiniMax, it is built for open-weight 428B MoE (~23B active per token) with a 1M-token context, native multimodal input — text, image and video, reports 59.0 on SWE-Bench Pro (a strong open-weight score), surpassing GPT-5.5, and low entry pricing at $0.30/$1.20 per million up to 512K tokens.
Its trade-offs: price doubles to $0.60/$2.40 above 512K tokens — not flat across 1M, miniMax's own reported 80.5 SWE-Bench Verified figure is vendor-stated, not independently verified, sWE-Bench Pro is a different, harder benchmark than SWE-Bench Verified, and newer than M2.7 but with less independent testing so far. At $0.3 in / $1.2 out per million tokens, it sits in the budget price band.
The bottom line for this matchup
Because MiniMax M2.7 and MiniMax M3 come from the same lab (MiniMax), they share the same training philosophy and ecosystem — the decision is purely tier vs. cost. MiniMax M3 is the more capable, more recent option; the other earns its place only when its price or latency profile fits a specific job better. Most teams should default to MiniMax M3 and drop down only with a concrete reason.
Frequently asked questions
Is MiniMax M2.7 or MiniMax M3 better for coding?
Public SWE-Bench figures are not available for either model, so the honest test is your own repository — run an identical real bug through both. By design, MiniMax M2.7 leans toward agentic and terminal coding well above its price tier (57.0 on terminal-bench 2, vendor-reported) while MiniMax M3 leans toward open-weight 428b moe (~23b active per token) with a 1m-token context, and that positioning usually predicts which feels better on your codebase.
Which is cheaper, MiniMax M2.7 or MiniMax M3?
They are priced almost identically, so cost will not decide between them.
Which has the bigger context window?
MiniMax M3 — 1M vs 205K, about 5.1× larger. Useful only if the model actually reasons over the full window, which not all do.
Should I upgrade from MiniMax M2.7 to MiniMax M3?
Since both are MiniMax models, the newer one (MiniMax M3) is usually the better default unless you need a specific cost or latency profile from the other.
Which is newer, MiniMax M2.7 or MiniMax M3?
MiniMax M3 — released June 2026, about 3 months after MiniMax M2.7.
MiniMax M2.7 vs MiniMax M3
MiniMax · China | MiniMax · China · Updated June 2026
Quick verdict
Both are MiniMax models. MiniMax M3 is the newer, generally stronger default; reach for MiniMax M2.7 when a specific cost or latency profile matters more than the latest capabilities.
MiniMax M2.7 and MiniMax M3 are both MiniMax models, so the real question is not which lab to trust but which tier fits your workload and budget. MiniMax M2.7 is a cheap open-weight agentic coder with near-frontier terminal scores — held back by a non-commercial licence and non-standard benchmarks. MiniMax M3 is miniMax's open-weight 428B multimodal model with a 1M context — strong reported coding (59.0 SWE-Bench Pro) at low entry pricing. Since both come from the same lab, the comparison below focuses on the tier-and-cost trade-offs that actually separate them.
Key differences at a glance
▸Context window: MiniMax M3 holds 5.1× more — 1M (~1,573 pages) vs 205K (~307 pages). But effective recall usually fades long before the advertised ceiling, so the bigger number only helps if the model reasons over it.
▸Recency: MiniMax M3 is the newer model by about 3 months (released June 2026), usually meaning fresher training data and capabilities.
Side-by-side specs
Spec
MiniMax M2.7
MiniMax M3
Provider
MiniMax (China)
MiniMax (China)
Released
March 18, 2026
June 2026
Context window
205K (~307 pages)
1M (~1,573 pages)
Price (in/out)
$0.3/$1.2 per 1M tokens
$0.3/$1.2 per 1M tokens
Open weight?
Yes — self-hostable
Yes — self-hostable
Modalities
text, code
text, image, video, code
SWE-Bench Verified
Not published
Not published
MRCR v2 @ 1M
Not published
Not published
Who wins what
Agentic and terminal coding well above its price tier (57.0 on Terminal-Bench 2, vendor-reported)
MiniMax M2.7
MiniMax M3 is comparatively weak here — price doubles to $0.60/$2.40 above 512K tokens — not flat across 1M
Independently ranked 14th of 97 on the Artificial Analysis Intelligence Index
MiniMax M2.7
MiniMax M3 is comparatively weak here — miniMax's own reported 80.5 SWE-Bench Verified figure is vendor-stated, not independently verified
Sparse mixture-of-experts — roughly 230B total but only ~10B active, so it runs on local hardware
MiniMax M2.7
MiniMax M2.7 lists sparse mixture-of-experts — roughly 230B total but only ~10B active, so it runs on local hardware among its strengths; MiniMax M3 does not.
Open-weight 428B MoE (~23B active per token) with a 1M-token context
MiniMax M3
Its 1M window holds about 5.1× more than MiniMax M2.7's 205K in a single prompt.
Native multimodal input — text, image and video
MiniMax M3
MiniMax's open-weight 428B multimodal model with a 1M context — strong reported coding (59.0 SWE-Bench Pro) at low entry pricing — and it carries the larger 1M context.
Reports 59.0 on SWE-Bench Pro (a strong open-weight score), surpassing GPT-5.5
MiniMax M3
MiniMax M2.7 is comparatively weak here — open weights but a NON-COMMERCIAL licence — commercial use requires prior written authorisation from MiniMax, and at least one major tracker still mislabels it as MIT
Largest single-prompt input
MiniMax M3
Its 1M window is about 5.1× larger than MiniMax M2.7's 205K, fitting roughly 1,573 pages in one prompt.
Which should you pick?
Someone analysing very long documents or codebases
→ MiniMax M3
Larger 1M window fits more in one prompt.
Anyone whose priority is agentic and terminal coding well above its price tier (57.0 on terminal-bench 2, vendor-reported)
→ MiniMax M2.7
It is specifically built for that.
Anyone whose priority is open-weight 428b moe (~23b active per token) with a 1m-token context
→ MiniMax M3
That is its strongest area.
MiniMax M2.7: where it fits
A cheap open-weight agentic coder with near-frontier terminal scores — held back by a non-commercial licence and non-standard benchmarks. Released March 18, 2026 by MiniMax, it is built for agentic and terminal coding well above its price tier (57.0 on Terminal-Bench 2, vendor-reported), independently ranked 14th of 97 on the Artificial Analysis Intelligence Index, sparse mixture-of-experts — roughly 230B total but only ~10B active, so it runs on local hardware, and served by five separate hosts at uniform pricing, so there is no provider lock-in.
Its trade-offs are real: open weights but a NON-COMMERCIAL licence — commercial use requires prior written authorisation from MiniMax, and at least one major tracker still mislabels it as MIT, reports SWE-Bench Pro instead of the standard Verified set, which blocks like-for-like comparison, and already superseded internally by M3, and its 205K context is small against 1M-class rivals. At $0.3 in / $1.2 out per million tokens, it sits in the budget price band.
MiniMax M3: where it fits
MiniMax's open-weight 428B multimodal model with a 1M context — strong reported coding (59.0 SWE-Bench Pro) at low entry pricing. Released June 2026 by MiniMax, it is built for open-weight 428B MoE (~23B active per token) with a 1M-token context, native multimodal input — text, image and video, reports 59.0 on SWE-Bench Pro (a strong open-weight score), surpassing GPT-5.5, and low entry pricing at $0.30/$1.20 per million up to 512K tokens.
Its trade-offs: price doubles to $0.60/$2.40 above 512K tokens — not flat across 1M, miniMax's own reported 80.5 SWE-Bench Verified figure is vendor-stated, not independently verified, sWE-Bench Pro is a different, harder benchmark than SWE-Bench Verified, and newer than M2.7 but with less independent testing so far. At $0.3 in / $1.2 out per million tokens, it sits in the budget price band.
The bottom line for this matchup
Because MiniMax M2.7 and MiniMax M3 come from the same lab (MiniMax), they share the same training philosophy and ecosystem — the decision is purely tier vs. cost. MiniMax M3 is the more capable, more recent option; the other earns its place only when its price or latency profile fits a specific job better. Most teams should default to MiniMax M3 and drop down only with a concrete reason.
Want both MiniMax M2.7 and MiniMax M3 without two subscriptions? LumiChats gives you these plus 40+ models under one ₹69/day pass (about $1/day) — draft with one, cross-check with the other.
Public SWE-Bench figures are not available for either model, so the honest test is your own repository — run an identical real bug through both. By design, MiniMax M2.7 leans toward agentic and terminal coding well above its price tier (57.0 on terminal-bench 2, vendor-reported) while MiniMax M3 leans toward open-weight 428b moe (~23b active per token) with a 1m-token context, and that positioning usually predicts which feels better on your codebase.
Which is cheaper, MiniMax M2.7 or MiniMax M3?
They are priced almost identically, so cost will not decide between them.
Which has the bigger context window?
MiniMax M3 — 1M vs 205K, about 5.1× larger. Useful only if the model actually reasons over the full window, which not all do.
Should I upgrade from MiniMax M2.7 to MiniMax M3?
Since both are MiniMax models, the newer one (MiniMax M3) is usually the better default unless you need a specific cost or latency profile from the other.
Which is newer, MiniMax M2.7 or MiniMax M3?
MiniMax M3 — released June 2026, about 3 months after MiniMax M2.7.
Specifications and benchmarks reflect publicly reported figures as of June 2026 and may change as providers release updates. Always verify on your own workload.