Muse Wonder vs Gemini 4 Argon
Released 24 hours apart: should you pay $4/$20 for Meta's Muse Wonder API or $2/$10 (until Nov 30) for Google's Gemini 4 Argon?
The short answer
| At a glance | Muse Wonder | Gemini 4 Argon |
|---|---|---|
| List priceinput / output per 1M tokens | $4 / $20 | $2 / $10 (better) |
| Benchmark task costArtificial Analysis, per Intelligence-Index task | $4.52 | $1.99 (better) |
| Intelligence IndexAA v4.3.2, higher is better | 52 | 53 (better) |
Meta's Muse Wonder (muse-wonder-1.0, launched Oct 12, 2026) and Google's Gemini 4 Argon (gemini-4-argon, launched Oct 13, 2026) landed 24 hours apart with converging long-term pricing: starting December 1, 2026, both charge the exact same $4.00 input / $20.00 output / $0.20 cached-read rate across a flat 1M context window with zero cache-write fees. Through November 30, 2026, however, Gemini 4 Argon's introductory $2.00 / $10.00 ($0.10 cached) pricing makes it half the per-token price and 2.3x cheaper per measured Artificial Analysis Intelligence-Index task ($1.99 vs $4.52), plus Argon offers 50%-off Batch and Flex tiers ($1/$5) that Meta does not publish for Muse Wonder. On same-harness AA v4.3.2 benchmarks, the two models split by workload: Muse Wonder decisively wins agentic knowledge-work Elo evaluations — GDPval-AA (1770 vs 1624, +146 Elo) and AA-Briefcase (1659 vs 1487, +172 Elo) — as well as professional PDF reasoning (GDP.pdf 31% vs 22%), whereas Gemini 4 Argon edges overall Intelligence Index (53 vs 52) and leads AutomationBench-AA (78% vs 68%), Terminal-Bench 4.0 (57% vs 52%), SciCode (62% vs 60%), and Humanity's Last Exam (57% vs 50%).
Choose Muse Wonder
Agentic knowledge work and document reasoning where Muse Wonder beats Argon in the same AA harness — GDPval-AA (1770 vs 1624 Elo), AA-Briefcase (1659 vs 1487 Elo), and GDP.pdf (31% vs 22%) — or long-horizon repo engineering where Meta's DeepSWE v1.1 (78.8% vs Argon's 77.9% Model Card) and MemoryBench (85.6%) shine.
Choose Gemini 4 Argon
2.3x lower per-task cost through November 30 ($1.99 vs $4.52 on AA), 50%-off Batch and Flex tiers ($1/$5 now, $2/$10 after Dec 1), SaaS workflow automation (AutomationBench-AA 78% vs 68%), terminal & scientific coding (Terminal-Bench 4.0 57% vs 52%, SciCode 62% vs 60%, HLE 57% vs 50%), or long-decode outputs up to 1M tokens.
Either — decide by evidence you generate
Post-December 1 standard-tier caching workloads: once Argon's intro window ends on 2026-12-01, both models bill $4.00 input / $20.00 output / $0.20 cached reads with no cache-write fee and no long-context cliff.
Key facts
- Through 2026-11-30, Gemini 4 Argon is half of Muse Wonder's list price ($2.00/$10.00/$0.10 vs $4.00/$20.00/$0.20 per MTok); starting 2026-12-01, both models charge the exact same $4.00/$20.00/$0.20 standard rates.
- Measured per-task cost (AA Intelligence Index v4.3.2): Gemini 4 Argon (High) costs $1.99/task (62k output tokens/task at $2/$10 intro rates) vs Muse Wonder (Max) at $4.52/task (54k output tokens/task at $4/$20). Note that at post-Nov-30 $4/$20 rates, Argon's 62k output tokens/task would cost ~$3.98 vs Muse Wonder's $4.52.
- Both models use 95%-off prefix caching with zero cache-write fees (cacheWriteRate = null) and flat pricing across a 1M context window.
- Service-tier availability: Gemini 4 Argon publishes Batch (-50%), Flex (-50%), and Priority (1.8x) tiers; Meta Model API publishes only Standard on-demand pricing for muse-wonder-1.0 (Contributor mode is excluded on API, and Muse Wonder Full is subscription-only in Muse Code).
- Same-harness AA v4.3.2 split: Muse Wonder leads GDPval-AA (1770 vs 1624 Elo), AA-Briefcase (1659 vs 1487 Elo), and GDP.pdf (31% vs 22%); Gemini 4 Argon leads overall AA Index (53 vs 52), AutomationBench-AA (78% vs 68%), Terminal-Bench 4.0 (57% vs 52%), SciCode (62% vs 60%), and HLE (57% vs 50%).
Official pricing, every tier
USD per 1M tokens from vendor pricing pages. 'Not published' means exactly that — we don't extrapolate.
| Service tier | Muse Wonder | Gemini 4 Argon |
|---|---|---|
| Standarddefault on-demand serving | in $4 · cached $0.20 · out $20 implicit cache only (no write fee) | in $2 · cached $0.10 · out $10 implicit cache only (no write fee) |
| Batchasync jobs, 50% off | Not published | in $1 · cached $0.05 · out $5 implicit cache only (no write fee) |
| Flexslower/cheaper serving, 50% off | Not published | in $1 · cached $0.05 · out $5 implicit cache only (no write fee) |
| Long-context band | Muse Wonder | Gemini 4 Argon |
|---|---|---|
| Long-context ratesapplies per request above the model threshold | Flat pricing same rates at any input size (up to context window) | Flat pricing same rates at any input size (up to context window) |
USD per 1M tokens. Every input token is billed in exactly one bucket — uncached, cached read, or cache write; a write replaces the input rate for those tokens (official rule), so write rates are per tier as listed. Google Gemini 4 Argon shows introductory rates valid through 2026-11-30 ($4/$20 standard starting 2026-12-01) and uses implicit caching only (95% discount on cache hits, no explicit write fee). · Google Gemini pricing (verified 2026-10-13) · Meta Muse Wonder (muse-wonder-1.0) is billed in Standard mode only ($4/$20, $0.20 cached reads, flat 1M context); Contributor mode is not available for Muse Wonder on Model API, and Muse Wonder Full is exclusive to Muse Code subscriptions. · Meta Model API pricing (verified 2026-10-13)
Tier ≠ quality change
Run this comparison on your own numbers
Preloaded with a workload typical for this comparison. Change anything — results recompute locally.
Token-based workload estimate only; it does not measure quality or cost per successful task.
| Model | Per request | Monthly |
|---|---|---|
| Muse Wonder | $0.214 | $2,140.00 |
| Gemini 4 Argon cheapest | $0.107 | $1,070.00 |
Monthly spread between the cheapest and priciest selected model: $1,070.00.
Breakdown: Muse Wonder · standard tier · standard context
- Fresh input (uncached): 30,000 tok × $4/M$0.12
- Cached input (reads): 70,000 tok × $0.2/M$0.014
- Cache writes: 0 tok × $0/M$0.00
- Output (incl. reasoning tokens): 4,000 tok × $20/M$0.08
- Per request$0.214
Rates: official pricing · verified 2026-10-13
- Per-token arithmetic only: it does not model quality differences or cross-vendor output-token-count differences for the same task.
- Reasoning tokens are billed as output tokens at the chosen model's output rate.
- Long-context band is decided per request by total input tokens vs the provider threshold, not by monthly volume.
- Cache accounting uses the official mutually-exclusive buckets: each input token is billed once, as uncached input, cached read, OR cache write — writes replace the input rate for those tokens, never add to it.
- Claude rows bill the 5-minute-TTL cache write; the published 1-hour TTL ($8/MTok standard) is not modeled (no TTL input).
- Rows flagged 'derived' are computed from a stated official rule (e.g. Anthropic batch -50%), not read verbatim from a price table.
- Missing official prices render as Unsupported — we never apply a guessed multiplier (e.g., GPT-6 Sol has no Ultrafast tier).
What the evidence actually says
Independent and vendor-run evals, each with publisher, runner, and harness. Numbers from different runs are never merged.
| Benchmark | Score | Effort | Harness | Evidence |
|---|---|---|---|---|
| AA Intelligence Indexv4.3.2 | ||||
| gemini-4-argon | 53 index-points | config 'Gemini 4 Argon (High)' per AA | Artificial Analysis Intelligence Index v4.3.2 (independent, same harness for all models) | Third-party |
| muse-wonder-1.0 | 52 index-points | config 'Muse Wonder (Max)' per AA | Artificial Analysis Intelligence Index v4.3.2 (independent, same harness across models) | Third-party |
| Terminal-Bench4.0 | ||||
| gemini-4-argon | 57 % | — | Artificial Analysis Intelligence Index v4.3.2 (independent, same harness for all models) | Third-party |
| muse-wonder-1.0 | 52 % | config 'Muse Wonder (Max)' per AA | Artificial Analysis Intelligence Index v4.3.2 (independent, same harness across models) | Third-party |
| SciCodeunspecified | ||||
| gemini-4-argon | 62 % | — | Artificial Analysis Intelligence Index v4.3.2 (independent, same harness for all models) | Third-party |
| muse-wonder-1.0 | 60 % | config 'Muse Wonder (Max)' per AA | Artificial Analysis Intelligence Index v4.3.2 (independent, same harness across models) | Third-party |
| GDPval-AAv2.1 | ||||
| gemini-4-argon | 1624 elo | — | Artificial Analysis Intelligence Index v4.3.2 (independent, same harness for all models) | Third-party |
| muse-wonder-1.0 | 1770 elo | config 'Muse Wonder (Max)' per AA | Artificial Analysis Intelligence Index v4.3.2 (independent, same harness across models) | Third-party |
| AA-Briefcasev1.1 | ||||
| gemini-4-argon | 1487 elo | — | Artificial Analysis Intelligence Index v4.3.2 (independent, same harness for all models) | Third-party |
| muse-wonder-1.0 | 1659 elo | config 'Muse Wonder (Max)' per AA | Artificial Analysis Intelligence Index v4.3.2 (independent, same harness across models) | Third-party |
| Humanity's Last Examunspecified | ||||
| gemini-4-argon | 57 % | — | Artificial Analysis Intelligence Index v4.3.2 (independent, same harness for all models) | Third-party |
| muse-wonder-1.0 | 50 % | config 'Muse Wonder (Max)' per AA | Artificial Analysis Intelligence Index v4.3.2 (independent, same harness across models) | Third-party |
| AutomationBench-AAunspecified | ||||
| gemini-4-argon | 78 % | — | Artificial Analysis Intelligence Index v4.3.2 (independent, same harness for all models) | Third-party |
| muse-wonder-1.0 | 68 % | config 'Muse Wonder (Max)' per AA | Artificial Analysis Intelligence Index v4.3.2 (independent, same harness across models) | Third-party |
| AA cost per Intelligence-Index taskv4.3.2 | ||||
| gemini-4-argon | 1.99 usd-per-task | — | Artificial Analysis Intelligence Index v4.3.2 (independent, same harness for all models) | Third-party |
| muse-wonder-1.0 | 4.52 usd-per-task | config 'Muse Wonder (Max)' per AA | Artificial Analysis Intelligence Index v4.3.2 (independent, same harness across models) | Third-party |
| Vals Index (DeepMind table)Oct 2026 | ||||
| gemini-4-argon | 68.9 % | — | Google DeepMind Model Card evaluation table (finance, legal, coding, and tax domains) | Vendor-reported |
| GraphWalks (256K–1M long context)BFS F1 | ||||
| gemini-4-argon | 84.2 % | — | Google DeepMind Model Card evaluation table | Vendor-reported |
| LABBench 2v2 | ||||
| gemini-4-argon | 88.8 % | — | Google DeepMind Model Card evaluation table | Vendor-reported |
| GDP.pdfunspecified | ||||
| muse-wonder-1.0 | 31 % | config 'Muse Wonder (Max)' per AA | Artificial Analysis Intelligence Index v4.3.2 (independent, same harness) | Third-party |
| AA Time to First Answer Tokenv4.3.2 | ||||
| muse-wonder-1.0 | 60.82 seconds | config 'Muse Wonder (Max)' per AA | Artificial Analysis Intelligence Index v4.3.2 latency test | Third-party |
| DeepSWEv1.1 | ||||
| muse-wonder-1.0 | 78.8 % | max | Meta Muse Wonder announcement evaluation table | Vendor-reported |
| SWE-Marathonunspecified | ||||
| muse-wonder-1.0 | 53.1 % | max | Meta Muse Wonder announcement evaluation table | Vendor-reported |
| MemoryBenchunspecified | ||||
| muse-wonder-1.0 | 85.6 % | max | Meta Muse Wonder announcement evaluation table | Vendor-reported |
| Harvey LAB-AAunspecified | ||||
| muse-wonder-1.0 | 10 % | max | Meta Muse Wonder announcement evaluation table | Vendor-reported |
Each row records who published and who ran it (see the Harness column); numbers compare only within a documented run.
Sources: Artificial Analysis (independent) — Independent same-harness comparison: GPT-6.1 Sol vs Claude Opus 5.5 (Intelligence Index v4.3.2) · Google DeepMind — Gemini 4 Argon Model Card (October 2026 evaluation table) · Artificial Analysis (independent) — Independent same-harness comparison: Muse Wonder (Max) vs Claude Opus 5.5 (Intelligence Index v4.3.2) · Meta — Put your ideas to work with Muse Wonder and Muse Wonder Full (official release blog & benchmark table)
Which model for which workload
Confidence labels: Measured = same-harness evidence · Inferred = reasoned from documented evidence · Insufficient evidence = no comparable public data.
| Workload | Our pick | Confidence | Why |
|---|---|---|---|
| Daily coding tasks | Gemini 4 Argon | Measured | 2.3x cheaper per AA task through Nov 30 ($1.99 vs $4.52) while leading same-harness Terminal-Bench 4.0 (57% vs 52%) and SciCode (62% vs 60%). |
| Complex repository maintenance | Muse Wonder | Measured | Scores 78.8% on DeepSWE v1.1 and 53.1% on SWE-Marathon in Meta's announcement table and slightly fewer output tokens per task (54k vs 62k) once both models sit at $4/$20 after Dec 1. DeepSWE (78.8% Meta table vs 77.9% DeepMind Model Card) comes from separate vendor publications. |
| Agent workflows | Muse Wonder | Measured | While Argon wins AutomationBench-AA (78% vs 68%), Muse Wonder leads real-world knowledge-work Elo by +146 on GDPval-AA (1770 vs 1624) and +172 on AA-Briefcase (1659 vs 1487) in the same AA harness. |
| Long-context tasks | Gemini 4 Argon | Price-derived | Both offer flat 1M context and zero cache-write fees, but Argon costs half as much per uncached and cached token ($2/$10/$0.10 vs $4/$20/$0.20) through Nov 30 and supports up to 1M output decode. |
| Quality-first | Gemini 4 Argon | Measured | Edges Muse Wonder on overall AA Intelligence Index v4.3.2 (53 vs 52) and Humanity's Last Exam (57% vs 50%), though Muse Wonder wins GDPval-AA and AA-Briefcase. |
| Budget-sensitive / high volume | Gemini 4 Argon | Measured | $1.99 vs $4.52 per AA task during the introductory window through Nov 30, plus 50%-off Batch and Flex tiers. |
| Offline batch jobs | Gemini 4 Argon | Price-derived | Argon publishes Batch and Flex tiers at $1.00/$5.00 ($0.05 cached) through Nov 30 ($2.00/$10.00 from Dec 1); Muse Wonder has no published Batch or Flex tier on Meta Model API. |
| Latency-sensitive production | Muse Wonder | Measured | Muse Wonder has a published AA Time to First Answer Token of 60.82s (37.41s TTFT) and 85–115 tok/s output speed on Meta Model API. |
Frequently asked questions
Is Muse Wonder a Google Gemini model or a Meta model?
Muse Wonder is a Meta model, not a Google Gemini model. Meta released Muse Wonder (muse-wonder-1.0) and Muse Wonder Full on October 12, 2026 via the Meta Model API (dev.meta.ai) and Muse Code. Google released Gemini 4 Argon (gemini-4-argon) the next day, October 13, 2026, via the Gemini API and Google AI Studio.
How do Muse Wonder and Gemini 4 Argon compare on API pricing?
Through November 30, 2026, Gemini 4 Argon is 50% cheaper on list rates: $2.00 input / $10.00 output / $0.10 cached read per 1M tokens (plus $1.00 / $5.00 Batch and Flex tiers), compared with Muse Wonder's $4.00 input / $20.00 output / $0.20 cached read on the Meta Model API. Starting December 1, 2026, Gemini 4 Argon's introductory window ends and its standard price steps up to $4.00 / $20.00 / $0.20 — matching Muse Wonder's exact standard rates.
Which model is better on independent benchmarks: Muse Wonder or Gemini 4 Argon?
On Artificial Analysis's same-harness Intelligence Index v4.3.2, Gemini 4 Argon (High) scores 53 overall vs 52 for API Muse Wonder (Max) (while subscription-only Muse Wonder Full scores 56). Argon leads on AutomationBench-AA (78% vs 68%), Terminal-Bench 4.0 (57% vs 52%), SciCode (62% vs 60%), and Humanity's Last Exam (57% vs 50%). However, Muse Wonder leads decisively on agentic knowledge-work Elo evaluations — GDPval-AA (1770 vs 1624) and AA-Briefcase (1659 vs 1487) — as well as PDF document reasoning (GDP.pdf 31% vs 22%).
Related
Sources & freshness
- Meta — Pricing and rate limits - Meta Model API (muse-wonder-1.0 standard pricing, caching, contributor mode)https://dev.meta.ai/docs/pricing-rate-limits · accessed 2026-10-13Muse Wonder (muse-wonder-1.0) Standard mode: $4.00 / 1M input tokens, $20.00 / 1M output tokens, $0.20 / 1M cached input tokens (95% discount on prefix cache hits; no explicit cache-write fee published; flat across 1,000,000-token context window). Contributor mode (discounted data-sharing mode available on muse-spark-1.3 at $0.10/$0.20) is explicitly NOT available for muse-wonder-1.0 on Model API (available only in Muse Code subscriptions). Rate limits: 3,000 RPM, 4,000,000 TPM per team. Web search grounding: $2.50 / 1,000 queries.
- Google — Gemini 4 Argon API model documentation (gemini-4-argon specs, 1M unified context, long decode)https://ai.google.dev/gemini-api/docs/models/gemini-4-argon · published 2026-10-13 · accessed 2026-10-13Model code gemini-4-argon: unified 1,048,576 input+output token window, max output 65,536 synchronous default or up to 1,048,576 with background long decode. Thinking levels low/medium/high (default medium). Implicit caching only (95% discount above 4,096 tokens).
- Google — Gemini Developer API pricing (Gemini 4 Argon standard/batch/flex/priority, context caching)https://ai.google.dev/gemini-api/docs/pricing · accessed 2026-10-13Gemini 4 Argon (gemini-4-argon) paid tier introductory pricing through 2026-11-30: Standard $2.00 input / $10.00 output / $0.10 context caching per 1M tokens ($4.00 / $20.00 / $0.20 starting 2026-12-01). Batch and Flex are 50% off ($1.00 / $5.00 / $0.05 through 2026-11-30; $2.00 / $10.00 / $0.10 starting 2026-12-01). Priority tier is 1.8x ($3.60 / $18.00 / $0.18 through 2026-11-30). Implicit caching only (explicit caching is not supported; no cache-write fee). Not available on the Free tier.
- Artificial Analysis (independent) — Independent same-harness comparison: GPT-6.1 Sol vs Claude Opus 5.5 (Intelligence Index v4.3.2)https://artificialanalysis.ai/models/comparisons/gpt-6-1-sol-vs-claude-opus-5-5 · accessed 2026-10-10Independent same-harness evaluation of both models. Cost-per-task = weighted average per Intelligence Index task; reflects measured token verbosity. Effort configs as labeled by AA ('Sol (Max)' vs 'Opus 5.5 (Max, Default Fallback)').
- Google DeepMind — Gemini 4 Argon Model Card (October 2026 evaluation table)https://deepmind.google/models/model-cards/gemini-4-argon/ · published 2026-10-13 · accessed 2026-10-13Official evaluation table comparing Gemini 4 Argon, GPT-6 Astra, Claude Fable 5.1, and Claude Opus 5.5 across Vals Index (68.9% vs 67.0% Opus), DeepSWE v1.1 (77.9% vs 74.2% Opus), Terminal-Bench 4.0 (57.4% vs 66.4% Opus), GraphWalks 256k-1M (84.2% vs 66.8% Opus), LABBench 2 (88.8% vs 73.1% Opus), and RiemannBench (76.0% vs 69.6% Opus). Knowledge cutoff March 2026.
- Artificial Analysis (independent) — Independent same-harness comparison: Muse Wonder (Max) vs Claude Opus 5.5 (Intelligence Index v4.3.2)https://artificialanalysis.ai/models/comparisons/muse-wonder-vs-claude-opus-5-5 · published 2026-10-12 · accessed 2026-10-13Independent same-harness Intelligence Index v4.3.2 evaluation of Muse Wonder (Max): Index 52 (vs 58 Opus 5.5, 53 Gemini 4 Argon, 52 GPT-6.1 Sol; subscription-only Muse Wonder Full scores 56), cost per Intelligence-Index task $4.52 (54k output tokens/task, 134M output tokens to run Index, $5,733 total) vs $5.98 Opus 5.5 (119k output tokens/task, 260M total), output speed 85 tok/s, time to first answer token 60.82s vs 748.61s Opus 5.5.
- Meta — Put your ideas to work with Muse Wonder and Muse Wonder Full (official release blog & benchmark table)https://dev.meta.ai/resources/blog/build-with-muse-wonder · published 2026-10-12 · accessed 2026-10-13Announces Muse Wonder (muse-wonder-1.0 on Meta Model API and Muse Code) and Muse Wonder Full (exclusively in Muse Code Power/Maximum Usage subscriptions, no pay-as-you-go API). Official evaluation table compares Muse Wonder Full (max), Muse Wonder (max), Claude Fable 5.1 (max), GPT-6 Astra (max), and Claude Opus 5.5 (max) across DeepSWE 1.1 (78.8% Wonder vs 74.2% Opus), SWE-Marathon (53.1% vs 46.3%), MemoryBench (85.6% vs 81.5%), Harvey LAB-AA (10.0% vs 4.2%), Terminal-Bench 4.0-AA (52.0% vs 59.6%), OSWorld V2.1 partial (72.6% vs 81.8%), and GDP.pdf (30.8% vs 30.6%).
Pricing data last verified 2026-10-13. OpenAI ships roughly weekly — re-verify before making spend commitments. See the methodology page for update cadence and stale-data flags.