Apertus 70B vs GPT-5-Codex vs Qwen3 VL 235B A22B Thinking
Qwen3 VL 235B A22B Thinking comes out ahead, 48 to 42 and 33 on our weighted score, though Apertus 70B is 6% cheaper per token.
Swiss AI
Apertus 70B
33/100- ECI—
- Price$0.82 / $2.42
- Context66K
OpenAIGPT-5-Codex
42/100- ECI—
- Price$1.25 / $10.00
- Context400K
- Our pick
Alibaba (Qwen)Qwen3 VL 235B A22B Thinking
48/100- ECI—
- Price$0.40 / $4.00
- Context131K
Qwen3 VL 235B A22B Thinking is our pick
Qwen3 VL 235B A22B Thinking is the better all-round choice, scoring 48/100 against GPT-5-Codex (42) and Apertus 70B (33). GPT-5-Codex wins on context window. The score weighs price 50%, inputs & features 30%, context window 20%. None of the shared benchmarks cover every model here yet, so capability is left out of this verdict.
- CapabilityNot enough dataNo independent benchmark covers every model here yet
- Lowest priceApertus 70BApertus 70B $1.22 · Qwen3 VL 235B A22B Thinking $1.30 · GPT-5-Codex $3.44 per 1M tokens (3:1 blend)
- Longest contextGPT-5-CodexGPT-5-Codex 400,000 · Qwen3 VL 235B A22B Thinking 131,072 · Apertus 70B 65,536 tokens
- Widest inputsGPT-5-Codex and Qwen3 VL 235B A22B ThinkingApertus 70B: Text · GPT-5-Codex: Text, Images · Qwen3 VL 235B A22B Thinking: Text, Images
- Self-hostingApertus 70B and Qwen3 VL 235B A22B ThinkingPublishes downloadable weights (Apache-2.0)
| Measure | Weight | Apertus 70B | GPT-5-Codex | Qwen3 VL 235B A22B Thinking |
|---|---|---|---|---|
| Price | 50% | 46 | 24 | 44 |
| Inputs & features | 30% | 25 | 70 | 70 |
| Context window | 20% | 12 | 44 | 24 |
| Overall | 100% | 33/100 | 42/100 | 48/100 |
Left out because at least one model lacks the data: capability. The remaining weights were rescaled.
Every spec in one table
Highlighted cells lead their row. Dashes mean the data is not published.
| Specification | Apertus 70BSwiss AI | GPT-5-CodexOpenAI | Qwen3 VL 235B A22B ThinkingAlibaba (Qwen) |
|---|---|---|---|
| Capability | |||
| Capabilities Index (ECI) | — | — | — |
| ECI rank | — | — | — |
| Price per million tokens | |||
| Input | $0.82 | $1.25 | $0.40 (best) |
| Output | $2.42 (best) | $10.00 | $4.00 |
| Cached input | — | — | — |
| Blended (3:1) | $1.22 (best) | $3.44 | $1.30 |
| Long-context rate | Same rate | Same rate | Same rate |
| Price source | Median of 3 providers | Median of 3 providers | Median of 9 providers |
| Limits | |||
| Context window | 65,536 tokens | 400,000 tokens (best) | 131,072 tokens |
| Max output | 8,192 tokens | 128,000 tokens (best) | 32,768 tokens |
| Inputs and features | |||
| Text | Yes | Yes | Yes |
| Images | No | Yes | Yes |
| PDFs | No | No | No |
| Audio | No | No | No |
| Video | No | No | No |
| Reasoning | No | Yes | Yes |
| Tool calling | Yes | Yes | Yes |
| Structured output | No | Yes | Yes |
| Availability | |||
| Weights | OpenApache-2.0 | Proprietary | Open |
| API model ID | — | — | — |
| API providers | 3 | 3 | 9 (best) |
| Released | Sep 2, 2025 | Sep 15, 2025 | Sep 23, 2025 |
| Knowledge cutoff | Sep 2025 | Sep 30, 2024 | Mar 31, 2025 |
What would a month cost?
Enter your expected volume in millions of tokens. List prices only; caching and batch discounts would lower these.
- Apertus 70B$13.04
- GPT-5-Codex$32.50
- Qwen3 VL 235B A22B Thinking$12.00
Which should you choose?
Which is better: Apertus 70B, GPT-5-Codex or Qwen3 VL 235B A22B Thinking?
Qwen3 VL 235B A22B Thinking is the better all-round choice, scoring 48/100 against GPT-5-Codex (42) and Apertus 70B (33). GPT-5-Codex wins on context window. The score weighs price 50%, inputs & features 30%, context window 20%. None of the shared benchmarks cover every model here yet, so capability is left out of this verdict.
Which is cheaper, Apertus 70B, GPT-5-Codex or Qwen3 VL 235B A22B Thinking?
Apertus 70B is cheaper at $0.82 input / $2.42 output per million tokens (median across 3 API providers). Qwen3 VL 235B A22B Thinking costs $0.40 input / $4.00 output per million tokens (median across 9 API providers); GPT-5-Codex costs $1.25 input / $10.00 output per million tokens (median across 3 API providers). At a typical mix of three input tokens to one output token, that is $1.22 per million tokens for Apertus 70B versus $1.30 for Qwen3 VL 235B A22B Thinking (1.1× as much) and $3.44 for GPT-5-Codex (2.8× as much).
Which scores higher on benchmarks?
There is no independent benchmark that covers all three models yet. Apertus 70B has not been scored yet, GPT-5-Codex has not been scored yet and Qwen3 VL 235B A22B Thinking has not been scored yet.
Which is better for coding?
There are no published SWE-bench Verified results for Apertus 70B, GPT-5-Codex and Qwen3 VL 235B A22B Thinking yet, so there is no like-for-like coding score. Test both on a sample of your own tasks. All three support tool calling for agent workflows.
Which has the bigger context window?
GPT-5-Codex has the largest context window at 400,000 tokens, against 131,072 for Qwen3 VL 235B A22B Thinking and 65,536 for Apertus 70B. Maximum output per response: Apertus 70B up to 8,192, GPT-5-Codex up to 128,000, Qwen3 VL 235B A22B Thinking up to 32,768 tokens.
Which can read images, PDFs, audio or video?
Apertus 70B accepts text; GPT-5-Codex accepts text and images; Qwen3 VL 235B A22B Thinking accepts text and images. GPT-5-Codex handles the widest range of inputs.
Are any of these open source?
Apertus 70B and Qwen3 VL 235B A22B Thinking publishes its weights (Apache-2.0) and can be self-hosted; GPT-5-Codex is proprietary.
Which is newer?
Qwen3 VL 235B A22B Thinking is the newest, released Sep 23, 2025. GPT-5-Codex came out Sep 15, 2025; Apertus 70B came out Sep 2, 2025. Knowledge cutoff: Apertus 70B Sep 2025, GPT-5-Codex Sep 30, 2024, Qwen3 VL 235B A22B Thinking Mar 31, 2025.
How do you decide the winner?
Each model gets a 0–100 score on capability (50%, independent benchmark results); price (25%, blended price per million tokens (3 input : 1 output), log scale); inputs & features (15%, image, PDF, audio and video input, tool calling, structured output and reasoning); context window (10%, maximum tokens per request, log scale). Dimensions missing for any model are dropped and the remaining weights rescaled, so every model is judged on the same evidence. Specs and prices come from public model listings and the labs’ own API pages; capability scores come from independent benchmark runs. Data updated Oct 4, 2026.
