The Best Free Model on OpenRouter Right Now
MiniMax M3 is the current editorial top pick. GLM 5.2 is the manual backup. Change only when your input, context, or data-policy requirement makes the difference material.
Context scope: Context lengths are the values reported for the exact OpenRouter :free model IDs in this snapshot. They are not maximum-context claims for every deployment, paid endpoint, or direct-provider version of the underlying model.
How OpenRouter's 50 / 1,000-request limits workMiniMax M3
It combines the strongest broad-work balance in the current eligible pool: high public capability signals, 1.05M context reported for its current OpenRouter :free model ID, and text, image, and video input.
- Free endpoint
- Listed in snapshot
- Snapshot context
- 1.05M on current :free ID
- Input
- Text · Image · Video
- Data policy
- Provider policy applies
Known limitation: Choose the ZDR-listed manual backup when endpoint retention is a hard requirement.
The current Top 5
| Rank | Model | Score | Important difference |
|---|---|---|---|
| #1 | MiniMax M3minimax/minimax-m3:free | 85 | Broad editorial top pick: its current OpenRouter :free ID reports 1.05M context with text, image, and video input. |
| #2 | GLM 5.2z-ai/glm-5.2:free | 84 | Highest published intelligence, agentic, and coding indices in the eligible pool. |
| #3 | MiniMax M2.7minimax/minimax-m2.7:free | 51 | Strong public coding signal without the multimodal surface of M3. |
| #4 | Gemma 4 31Bgoogle/gemma-4-31b-it:free | 49 | Multimodal alternative whose current OpenRouter :free ID reports 262K context. |
| #5 | Gemma 4 26B A4Bgoogle/gemma-4-26b-a4b-it:free | 34 | Another multimodal choice when the exact Gemma variant matters. |
Choose another model only if…
- You require a ZDR-listed free endpoint. Choose GLM 5.2 in this snapshot.
- You need the highest published text-only capability signals. GLM 5.2 leads the eligible pool on all three public indices used here.
- Your selected endpoint is unavailable or rate-limited. Switch to the listed backup manually instead of rebuilding the Chat or its files.
- Your provider policy excludes the top endpoint. Treat data policy as a hard filter, not a few points in a capability score.
How scoring works
The score ranks concrete free endpoints for one broad editorial top pick. Capability requirements decide which models enter the eligible pool; the fixed formula then orders that pool.
For scoring, context means the value reported for each exact OpenRouter :free model ID in this dated snapshot. It does not claim that the underlying model family is capped at that value across other routes, paid endpoints, or direct deployments.
The eligible pool contains 6 models. For each benchmark, percentile equals the number of pool values strictly below the model divided by 5; tied values share a percentile. Each weighted benchmark component is rounded to the nearest integer before adding context and modality points. For those exact :free IDs, reported context contributes 6 / 9 / 12 / 15 points at 128K / 256K / 512K / 1M. Image input contributes six points; audio or video contributes four more. Equal totals sort by canonical model ID. Data policy is a separate decision filter, not a hidden score.
Complete score calculation
| Eligible model | Intelligence raw → percentile → points | Agentic raw → percentile → points | Coding raw → percentile → points | Snapshot context for exact :free ID | Modalities | Total |
|---|---|---|---|---|---|---|
Gemma 4 26B A4Bgoogle/gemma-4-26b-a4b-it:free | 26.1 → P20 → 8 | 11 → P20 → 5 | 39.3 → P20 → 2 | 262K on current :free ID → 9 | Text · Image · Video → 10 | 34 |
Gemma 4 31Bgoogle/gemma-4-31b-it:free | 29.7 → P40 → 16 | 14.4 → P40 → 10 | 43.4 → P40 → 4 | 262K on current :free ID → 9 | Text · Image · Video → 10 | 49 |
MiniMax M2.7minimax/minimax-m2.7:free | 38.9 → P60 → 24 | 25.9 → P60 → 15 | 52.6 → P60 → 6 | 197K on current :free ID → 6 | Text → 0 | 51 |
MiniMax M3minimax/minimax-m3:free | 45.4 → P80 → 32 | 36.1 → P80 → 20 | 58.6 → P80 → 8 | 1.05M on current :free ID → 15 | Text · Image · Video → 10 | 85 |
Nemotron 3 Supernvidia/nemotron-3-super-120b-a12b:free | 25.7 → P0 → 0 | 8.8 → P0 → 0 | 37.7 → P0 → 0 | 262K on current :free ID → 9 | Text → 0 | 9 |
GLM 5.2z-ai/glm-5.2:free | 52.6 → P100 → 40 | 45.7 → P100 → 25 | 68.8 → P100 → 10 | 256K on current :free ID → 9 | Text → 0 | 84 |
Query retrieved 2026-08-30T10:46:46Z. Models response SHA-256: 6711caf4019bcf0a37eb0095472c6b93005ba928677524799ee428a483637c33. ZDR response SHA-256: 4980c6778224dde6214bf89a76353830a71acd4f497222d4e54ac20f32453c95. Download the exact dated input snapshot.
Updates are manual: retrieve the two documented sources, freeze a new dated snapshot, reapply the same eligibility rules and formula, review every rank change, then rebuild the page. Ordinary monthly updates do not change the weights.
Eligibility before ranking
- Listed as a current :free model in the OpenRouter catalog snapshot
- Supports tools and tool choice
- Supports response_format or structured_outputs
- Provides at least 128K of free-endpoint context
- Includes all three published Artificial Analysis indices used by the formula