Pick GLM 4.7 for genuinely permissive open weights — an mit-licensed 358b mixture-of-experts with no commercial restrictions or strong agentic coding for the price — 73.8% on swe-bench verified undercut most closed frontier models at launch. Pick Grok 4.7 for 2.1 trillion parameters, up 40% from grok 4.6's 1.5 trillion, at the same $2/$6 per million token price or deepswe v1.1 (high effort): 71.0%, up from grok 4.6's 65.2%; cursorbench 4.0: 46.3%, up from 40.4%. Choose GLM 4.7 if you need self-hosting or data privacy; Grok 4.7 if you want a managed API.
GLM 4.7 (Z.ai, China) and Grok 4.7 (xAI, US) line up two different AI ecosystems against each other — a comparison that is as much about cost philosophy and openness as raw capability. GLM 4.7 is an MIT-licensed 358B open mixture-of-experts with strong 73.8% SWE-Bench Verified coding — but two generations behind GLM 5.2. Grok 4.7 is xAI's September 21, 2026 update — a 2.1T-parameter model trained partly on SpaceX hardware data, same price as Grok 4.6 but stronger on coding benchmarks. They diverge most on price, context window and open vs. closed weights — each quantified below from the models' real specs.
Key differences
Price: GLM 4.7 is about 3.3× cheaper on input ($0.6/$2.2 per 1M tokens vs $2/$6 per 1M tokens) — meaningful once you are processing millions of tokens a month.
Context window: Grok 4.7 holds 2.5× more — 500K tokens (~750 pages) vs 200K (~304 pages). But effective recall usually fades long before the advertised ceiling, so the bigger number only helps if the model reasons over it.
Recency: Grok 4.7 is the newer model by about 9 months (released September 21, 2026), usually meaning fresher training data and capabilities.
Ecosystem: this is a China-vs-US matchup — they differ in pricing philosophy, data-residency options, and tooling ecosystems, not only benchmarks.
Specifications
Spec
GLM 4.7
Grok 4.7
Provider
Z.ai (China)
xAI (US)
Released
December 22, 2025
September 21, 2026
Context window
200K (~304 pages)
500K tokens (~750 pages)
Price (in/out)
$0.6/$2.2 per 1M tokens
$2/$6 per 1M tokens
Open weight?
Yes — self-hostable
No — API only
Modalities
text, code
text, image, code
SWE-Bench Verified
73.8%
Not published
MRCR v2 @ 1M
Not published
Not published
Who wins what
Genuinely permissive open weights — an MIT-licensed 358B mixture-of-experts with no commercial restrictions: GLM 4.7 — Open weights make this possible at all — Grok 4.7 is API-only, so it cannot leave the vendor's servers.
Strong agentic coding for the price — 73.8% on SWE-Bench Verified undercut most closed frontier models at launch: GLM 4.7 — At $0.6/$2.2 per 1M tokens it undercuts Grok 4.7 ($2/$6 per 1M tokens), and that gap compounds at volume.
An unusually generous 128K maximum output, which suits bulk refactors and long generation: GLM 4.7 — An MIT-licensed 358B open mixture-of-experts with strong 73.8% SWE-Bench Verified coding — but two generations behind GLM 5.2 — and it runs cheaper at $0.6/$2.2 per 1M tokens.
2.1 trillion parameters, up 40% from Grok 4.6's 1.5 trillion, at the same $2/$6 per million token price: Grok 4.7 — Its 500K tokens window holds about 2.5× more than GLM 4.7's 200K in a single prompt.
DeepSWE v1.1 (high effort): 71.0%, up from Grok 4.6's 65.2%; CursorBench 4.0: 46.3%, up from 40.4%: Grok 4.7 — XAI's September 21, 2026 update — a 2.1T-parameter model trained partly on SpaceX hardware data, same price as Grok 4.6 but stronger on coding benchmarks — and it carries the larger 500K tokens context.
Trained with supplemental SpaceX data (Starlink telemetry, manufacturing records, engineering failure logs) — xAI says this improves reasoning about hardware and physical systems: Grok 4.7 — GLM 4.7 is comparatively weak here — text-only with no vision, and self-hosting a 358B model is a serious hardware commitment
Lowest cost at scale: GLM 4.7 — At $0.6/$2.2 per 1M tokens, it is the cheaper of the two — the gap dominates the bill on high-volume workloads.
Largest single-prompt input: Grok 4.7 — Its 500K tokens window is about 2.5× larger than GLM 4.7's 200K, fitting roughly 750 pages in one prompt.
Which should you pick?
A cost-sensitive startup shipping high volume: GLM 4.7 — At $0.6/$2.2 per 1M tokens it undercuts Grok 4.7, and on millions of tokens that margin decides the monthly bill.
Someone analysing very long documents or codebases: Grok 4.7 — Larger 500K tokens window fits more in one prompt.
A team with data-privacy or self-hosting needs: GLM 4.7 — Open weights let you run it on your own hardware; Grok 4.7 is API-only.
Anyone whose priority is genuinely permissive open weights — an mit-licensed 358b mixture-of-experts with no commercial restrictions: GLM 4.7 — It is specifically built for that.
Anyone whose priority is 2.1 trillion parameters, up 40% from grok 4.6's 1.5 trillion, at the same $2/$6 per million token price: Grok 4.7 — That is its strongest area.
An enterprise with regional data-residency rules: Grok 4.7 or GLM 4.7 — Origin (China vs US) affects where data is processed and which compliance regime applies — check the provider's terms for your region.
GLM 4.7: where it fits
An MIT-licensed 358B open mixture-of-experts with strong 73.8% SWE-Bench Verified coding — but two generations behind GLM 5.2. Released December 22, 2025 by Z.ai, it is built for genuinely permissive open weights — an MIT-licensed 358B mixture-of-experts with no commercial restrictions, strong agentic coding for the price — 73.8% on SWE-Bench Verified undercut most closed frontier models at launch, an unusually generous 128K maximum output, which suits bulk refactors and long generation, and cheap long-running agent loops thanks to aggressive prompt caching.
Its trade-offs are real: two generations behind — GLM 5, 5.1 and 5.2 have all shipped since, and new builds should default to those, its Verified lead narrows sharply on harder evaluations like SWE-Bench Pro, and text-only with no vision, and self-hosting a 358B model is a serious hardware commitment. At $0.6 in / $2.2 out per million tokens, it sits in the budget price band.
Grok 4.7: where it fits
XAI's September 21, 2026 update — a 2.1T-parameter model trained partly on SpaceX hardware data, same price as Grok 4.6 but stronger on coding benchmarks. Released September 21, 2026 by xAI, it is built for 2.1 trillion parameters, up 40% from Grok 4.6's 1.5 trillion, at the same $2/$6 per million token price, deepSWE v1.1 (high effort): 71.0%, up from Grok 4.6's 65.2%; CursorBench 4.0: 46.3%, up from 40.4%, trained with supplemental SpaceX data (Starlink telemetry, manufacturing records, engineering failure logs) — xAI says this improves reasoning about hardware and physical systems, and xAI's strongest safety guardrails to date, per the company.
Its trade-offs: release was delayed at least five times since late July 2026 before shipping, 500K context window trails several rivals now sitting at 1M+, and reviewers note it arrives "late to the AI frontier party" against GPT-6 Astra, Claude Fable 5.1 and Opus 5.5, all shipped in the weeks just before it. At $2 in / $6 out per million tokens, it sits in the mid price band.
The bottom line for this matchup
The defining split here is open vs. closed. GLM 4.7 gives you weights you control — self-host it, fine-tune it, keep data in-house, pay only for hardware. Grok 4.7 gives you a managed, always-updated API with no infrastructure to run. Teams with GPUs, privacy requirements, or huge volume often favour the open model; teams that want zero ops and the latest capabilities favour the closed one. Capability is close enough that this operational question, not the benchmark, usually decides it.
Frequently asked questions
Is GLM 4.7 or Grok 4.7 better for coding?
Public SWE-Bench figures are not available for Grok 4.7, so the honest test is your own repository — run an identical real bug through both. By design, GLM 4.7 leans toward genuinely permissive open weights — an mit-licensed 358b mixture-of-experts with no commercial restrictions while Grok 4.7 leans toward 2.1 trillion parameters, up 40% from grok 4.6's 1.5 trillion, at the same $2/$6 per million token price, and that positioning usually predicts which feels better on your codebase.
Which is cheaper, GLM 4.7 or Grok 4.7?
GLM 4.7 is open-weight, so self-hosting means no per-token fee (you pay for hardware instead), while Grok 4.7 is API-metered at $2/$6 per 1M tokens. For most teams without GPUs, the API model is cheaper to start; at very high volume, self-hosting can win.
Which has the bigger context window?
Grok 4.7 — 500K tokens vs 200K, about 2.5× larger. Useful only if the model actually reasons over the full window, which not all do.
Can I use both GLM 4.7 and Grok 4.7 together?
Yes — a multi-model platform like LumiChats gives you GLM 4.7, Grok 4.7 and 40+ others under one ₹69/day pass (about $1/day), so you can draft with one and cross-check with the other instead of buying two subscriptions.
Which is newer, GLM 4.7 or Grok 4.7?
Grok 4.7 — released September 21, 2026, about 9 months after GLM 4.7.
GLM 4.7 vs Grok 4.7
Z.ai · China | xAI · US · Updated June 2026
Quick verdict
Pick GLM 4.7 for genuinely permissive open weights — an mit-licensed 358b mixture-of-experts with no commercial restrictions or strong agentic coding for the price — 73.8% on swe-bench verified undercut most closed frontier models at launch. Pick Grok 4.7 for 2.1 trillion parameters, up 40% from grok 4.6's 1.5 trillion, at the same $2/$6 per million token price or deepswe v1.1 (high effort): 71.0%, up from grok 4.6's 65.2%; cursorbench 4.0: 46.3%, up from 40.4%. Choose GLM 4.7 if you need self-hosting or data privacy; Grok 4.7 if you want a managed API.
GLM 4.7 (Z.ai, China) and Grok 4.7 (xAI, US) line up two different AI ecosystems against each other — a comparison that is as much about cost philosophy and openness as raw capability. GLM 4.7 is an MIT-licensed 358B open mixture-of-experts with strong 73.8% SWE-Bench Verified coding — but two generations behind GLM 5.2. Grok 4.7 is xAI's September 21, 2026 update — a 2.1T-parameter model trained partly on SpaceX hardware data, same price as Grok 4.6 but stronger on coding benchmarks. They diverge most on price, context window and open vs. closed weights — each quantified below from the models' real specs.
Key differences at a glance
▸Price: GLM 4.7 is about 3.3× cheaper on input ($0.6/$2.2 per 1M tokens vs $2/$6 per 1M tokens) — meaningful once you are processing millions of tokens a month.
▸Context window: Grok 4.7 holds 2.5× more — 500K tokens (~750 pages) vs 200K (~304 pages). But effective recall usually fades long before the advertised ceiling, so the bigger number only helps if the model reasons over it.
▸Recency: Grok 4.7 is the newer model by about 9 months (released September 21, 2026), usually meaning fresher training data and capabilities.
▸Ecosystem: this is a China-vs-US matchup — they differ in pricing philosophy, data-residency options, and tooling ecosystems, not only benchmarks.
Side-by-side specs
Spec
GLM 4.7
Grok 4.7
Provider
Z.ai (China)
xAI (US)
Released
December 22, 2025
September 21, 2026
Context window
200K (~304 pages)
500K tokens (~750 pages)
Price (in/out)
$0.6/$2.2 per 1M tokens
$2/$6 per 1M tokens
Open weight?
Yes — self-hostable
No — API only
Modalities
text, code
text, image, code
SWE-Bench Verified
73.8%
Not published
MRCR v2 @ 1M
Not published
Not published
Who wins what
Genuinely permissive open weights — an MIT-licensed 358B mixture-of-experts with no commercial restrictions
GLM 4.7
Open weights make this possible at all — Grok 4.7 is API-only, so it cannot leave the vendor's servers.
Strong agentic coding for the price — 73.8% on SWE-Bench Verified undercut most closed frontier models at launch
GLM 4.7
At $0.6/$2.2 per 1M tokens it undercuts Grok 4.7 ($2/$6 per 1M tokens), and that gap compounds at volume.
An unusually generous 128K maximum output, which suits bulk refactors and long generation
GLM 4.7
An MIT-licensed 358B open mixture-of-experts with strong 73.8% SWE-Bench Verified coding — but two generations behind GLM 5.2 — and it runs cheaper at $0.6/$2.2 per 1M tokens.
2.1 trillion parameters, up 40% from Grok 4.6's 1.5 trillion, at the same $2/$6 per million token price
Grok 4.7
Its 500K tokens window holds about 2.5× more than GLM 4.7's 200K in a single prompt.
DeepSWE v1.1 (high effort): 71.0%, up from Grok 4.6's 65.2%; CursorBench 4.0: 46.3%, up from 40.4%
Grok 4.7
XAI's September 21, 2026 update — a 2.1T-parameter model trained partly on SpaceX hardware data, same price as Grok 4.6 but stronger on coding benchmarks — and it carries the larger 500K tokens context.
Trained with supplemental SpaceX data (Starlink telemetry, manufacturing records, engineering failure logs) — xAI says this improves reasoning about hardware and physical systems
Grok 4.7
GLM 4.7 is comparatively weak here — text-only with no vision, and self-hosting a 358B model is a serious hardware commitment
Lowest cost at scale
GLM 4.7
At $0.6/$2.2 per 1M tokens, it is the cheaper of the two — the gap dominates the bill on high-volume workloads.
Largest single-prompt input
Grok 4.7
Its 500K tokens window is about 2.5× larger than GLM 4.7's 200K, fitting roughly 750 pages in one prompt.
Which should you pick?
A cost-sensitive startup shipping high volume
→ GLM 4.7
At $0.6/$2.2 per 1M tokens it undercuts Grok 4.7, and on millions of tokens that margin decides the monthly bill.
Someone analysing very long documents or codebases
→ Grok 4.7
Larger 500K tokens window fits more in one prompt.
A team with data-privacy or self-hosting needs
→ GLM 4.7
Open weights let you run it on your own hardware; Grok 4.7 is API-only.
Anyone whose priority is genuinely permissive open weights — an mit-licensed 358b mixture-of-experts with no commercial restrictions
→ GLM 4.7
It is specifically built for that.
Anyone whose priority is 2.1 trillion parameters, up 40% from grok 4.6's 1.5 trillion, at the same $2/$6 per million token price
→ Grok 4.7
That is its strongest area.
An enterprise with regional data-residency rules
→ Grok 4.7 or GLM 4.7
Origin (China vs US) affects where data is processed and which compliance regime applies — check the provider's terms for your region.
GLM 4.7: where it fits
An MIT-licensed 358B open mixture-of-experts with strong 73.8% SWE-Bench Verified coding — but two generations behind GLM 5.2. Released December 22, 2025 by Z.ai, it is built for genuinely permissive open weights — an MIT-licensed 358B mixture-of-experts with no commercial restrictions, strong agentic coding for the price — 73.8% on SWE-Bench Verified undercut most closed frontier models at launch, an unusually generous 128K maximum output, which suits bulk refactors and long generation, and cheap long-running agent loops thanks to aggressive prompt caching.
Its trade-offs are real: two generations behind — GLM 5, 5.1 and 5.2 have all shipped since, and new builds should default to those, its Verified lead narrows sharply on harder evaluations like SWE-Bench Pro, and text-only with no vision, and self-hosting a 358B model is a serious hardware commitment. At $0.6 in / $2.2 out per million tokens, it sits in the budget price band.
Grok 4.7: where it fits
XAI's September 21, 2026 update — a 2.1T-parameter model trained partly on SpaceX hardware data, same price as Grok 4.6 but stronger on coding benchmarks. Released September 21, 2026 by xAI, it is built for 2.1 trillion parameters, up 40% from Grok 4.6's 1.5 trillion, at the same $2/$6 per million token price, deepSWE v1.1 (high effort): 71.0%, up from Grok 4.6's 65.2%; CursorBench 4.0: 46.3%, up from 40.4%, trained with supplemental SpaceX data (Starlink telemetry, manufacturing records, engineering failure logs) — xAI says this improves reasoning about hardware and physical systems, and xAI's strongest safety guardrails to date, per the company.
Its trade-offs: release was delayed at least five times since late July 2026 before shipping, 500K context window trails several rivals now sitting at 1M+, and reviewers note it arrives "late to the AI frontier party" against GPT-6 Astra, Claude Fable 5.1 and Opus 5.5, all shipped in the weeks just before it. At $2 in / $6 out per million tokens, it sits in the mid price band.
The bottom line for this matchup
The defining split here is open vs. closed. GLM 4.7 gives you weights you control — self-host it, fine-tune it, keep data in-house, pay only for hardware. Grok 4.7 gives you a managed, always-updated API with no infrastructure to run. Teams with GPUs, privacy requirements, or huge volume often favour the open model; teams that want zero ops and the latest capabilities favour the closed one. Capability is close enough that this operational question, not the benchmark, usually decides it.
Want both GLM 4.7 and Grok 4.7 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 Grok 4.7, so the honest test is your own repository — run an identical real bug through both. By design, GLM 4.7 leans toward genuinely permissive open weights — an mit-licensed 358b mixture-of-experts with no commercial restrictions while Grok 4.7 leans toward 2.1 trillion parameters, up 40% from grok 4.6's 1.5 trillion, at the same $2/$6 per million token price, and that positioning usually predicts which feels better on your codebase.
Which is cheaper, GLM 4.7 or Grok 4.7?
GLM 4.7 is open-weight, so self-hosting means no per-token fee (you pay for hardware instead), while Grok 4.7 is API-metered at $2/$6 per 1M tokens. For most teams without GPUs, the API model is cheaper to start; at very high volume, self-hosting can win.
Which has the bigger context window?
Grok 4.7 — 500K tokens vs 200K, about 2.5× larger. Useful only if the model actually reasons over the full window, which not all do.
Can I use both GLM 4.7 and Grok 4.7 together?
Yes — a multi-model platform like LumiChats gives you GLM 4.7, Grok 4.7 and 40+ others under one ₹69/day pass (about $1/day), so you can draft with one and cross-check with the other instead of buying two subscriptions.
Which is newer, GLM 4.7 or Grok 4.7?
Grok 4.7 — released September 21, 2026, about 9 months after GLM 4.7.
Specifications and benchmarks reflect publicly reported figures as of June 2026 and may change as providers release updates. Always verify on your own workload.