DeepSeek V4 Flash Vision Exp vs Qwen3.8 Flash Next vs Mercury 2.5
No independent capability benchmark covers every model here yet. The score uses price and context window.No independent capability benchmark covers every model here yet. The score uses context window.
DeepSeek
DeepSeek V4 Flash Vision Exp
68/100- ECI—
- Price$0.216 / $0.647
- Context1M
Alibaba (Qwen)
Qwen3.8 Flash Next
65/100- ECI—
- Price$0.20 / $0.50
- Context262K
Inception
Mercury 2.5
62/100- ECI—
- Price$0.20 / $0.75
- Context260K
DeepSeek
DeepSeek V4 Flash Vision Exp
64/100- ECI—
- Price$0.216 / $0.647
- Context1M
Alibaba (Qwen)
Qwen3.8 Flash Next
43/100- ECI—
- Price$0.20 / $0.50
- Context262K
Inception
Mercury 2.5
43/100- ECI—
- Price$0.20 / $0.75
- Context260K
Capability evidence is incomplete
There is not enough shared capability evidence to name an overall winner. The scores (DeepSeek V4 Flash Vision Exp 68/100, Qwen3.8 Flash Next 65/100, Mercury 2.5 62/100) compare the available specifications and price. The score weighs price 71%, context window 29%. None of the shared benchmarks cover every model here yet, so capability is left out of this verdict.There is not enough shared capability evidence to name an overall winner. The scores (DeepSeek V4 Flash Vision Exp 64/100, Qwen3.8 Flash Next 43/100, Mercury 2.5 43/100) compare the available specifications only. The score weighs context window 100%. 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 priceQwen3.8 Flash NextQwen3.8 Flash Next $0.275 · DeepSeek V4 Flash Vision Exp $0.323 · Mercury 2.5 $0.338 per 1M tokens (3:1 blend)
- Longest contextDeepSeek V4 Flash Vision ExpDeepSeek V4 Flash Vision Exp 1,000,000 · Qwen3.8 Flash Next 262,144 · Mercury 2.5 260,000 tokens
- Widest inputsQwen3.8 Flash NextDeepSeek V4 Flash Vision Exp: Text, Images · Qwen3.8 Flash Next: Text, Images, Video · Mercury 2.5: Text
- Self-hostingDeepSeek V4 Flash Vision Exp and Qwen3.8 Flash NextPublishes downloadable weights (MIT and qwen-community-1.0)
| Measure | Weight | DeepSeek V4 Flash Vision Exp | Qwen3.8 Flash Next | Mercury 2.5 |
|---|---|---|---|---|
| Price | 71% | 70 | 73 | 69 |
| Context window | 29% | 64 | 43 | 43 |
| Specs & price only | 100% | 68/100 | 65/100 | 62/100 |
Left out because at least one model lacks the data: capability and inputs & features. The remaining weights were rescaled.
| Measure | Weight | DeepSeek V4 Flash Vision Exp | Qwen3.8 Flash Next | Mercury 2.5 |
|---|---|---|---|---|
| Context window | 100% | 64 | 43 | 43 |
| Specs only | 100% | 64/100 | 43/100 | 43/100 |
Left out because at least one model lacks the data: capability and inputs & features. The remaining weights were rescaled.
Every spec in one table
Highlighted cells lead their row. Unavailable specifications are marked “Not published”. Publisher results include their evaluation settings and source.
| Specification | |||
|---|---|---|---|
| Capability | |||
| Capabilities Index (ECI) | Not published | Not published | Not published |
| ECI rank | Not published | Not published | Not published |
| ECI confidence interval | Not published | Not published | Not published |
| ECI evaluation date | Not published | Not published | Not published |
| Documented capabilities | Experimental multimodal DeepSeek V4 Flash model for image understanding, coding, and agentic work | Open-weight experimental preview of the Qwen4 architecture: hybrid-attention MoE (125B total, 6B active) with vision encoder for coding, agent tasks, and image and video understanding | Reasoning model for deliberate analysis, multi-step problem solving, and tool use |
| Model parameters | Not published | Not published | Not published |
| Output modalities | Text | Text | Text |
| Publisher-reported benchmarks | |||
| Terminal-Bench 2.1max effort | 83.9scoreReported result | Not published | Not published |
| NL2Repomax effort | 57.7scoreReported result | Not published | Not published |
| CyberGymmax effort | 75.3scoreReported result | Not published | Not published |
| DeepSWEmax effort | 59.3scoreReported result | Not published | Not published |
| Toolathlon-Verifiedmax effort | 75.9scoreReported result | Not published | Not published |
| DSBench-Hard · internalmax effort | 63.6scoreReported result | Not published | Not published |
| AutomationBench · publicmax effort | 25.7scoreReported result | Not published | Not published |
| ApexBenchPublisher-reported result | 36.5pass@1Reported result | Not published | Not published |
| Agents' Last ExamPublisher-reported result | 27.3scoreReported result | 51.2scoreReported result | Not published |
| ChartographyPublisher-reported result | 64.3scoreReported result | Not published | Not published |
| ZeroBenchPublisher-reported result | 35pass@5Reported result | Not published | Not published |
| DeepSWE 1.1Publisher-reported result | Not published | 58.7scoreReported result | Not published |
| SWE-Bench Pro · Qwen refined and corrected task setPublisher-reported result | Not published | 62.5scoreReported result | Not published |
| SWE-Bench MultilingualPublisher-reported result | Not published | 81scoreReported result | Not published |
| NL2RepoPublisher-reported result | Not published | 48.1scoreReported result | Not published |
| CoWorkBench · internalPublisher-reported result | Not published | 73.9scoreReported result | Not published |
| JobBenchPublisher-reported result | Not published | 55.7scoreReported result | Not published |
| Agents' Last ExamPublisher-reported result | Not published | 24.3pass@1Reported result | Not published |
| Toolathlon-VerifiedPublisher-reported result | Not published | 73.5pass@1Reported result | Not published |
| IFBenchPublisher-reported result | Not published | 81.3scoreReported result | Not published |
| GPQA DiamondPublisher-reported result | Not published | 91.7scoreReported result | Not published |
| Humanity's Last Examwithout tools; GPT-4o judge | Not published | 35.9scoreReported result | Not published |
| LiveCodeBench 6Publisher-reported result | Not published | 91.9scoreReported result | Not published |
| ClawEval-MMPublisher-reported result | Not published | 64.4pass@3Reported result | Not published |
| ClawEval-MMPublisher-reported result | Not published | 60.4average scoreReported result | Not published |
| RecreationBench · internalPublisher-reported result | Not published | 49.9scoreReported result | Not published |
| AndroidWorldPublisher-reported result | Not published | 84.5scoreReported result | Not published |
| OSWorld 2.0Publisher-reported result | Not published | 19.4binary completion rateReported result | Not published |
| OSWorld 2.0Publisher-reported result | Not published | 52.3partial scoreReported result | Not published |
| Vision2Webgpt-5.4-2026-03-05 judge | Not published | 64scoreReported result | Not published |
| ERQAPublisher-reported result | Not published | 72.3scoreReported result | Not published |
| LVBenchPublisher-reported result | Not published | 76.6scoreReported result | Not published |
| RealWorldQAPublisher-reported result | Not published | 88.5scoreReported result | Not published |
| MathVision · corrected annotationswithout code interpreter | Not published | 90.6scoreReported result | Not published |
| MathVision · corrected annotationswith code interpreter | Not published | 95.7scoreReported result | Not published |
| CharXiv Reasoningwithout code interpreter | Not published | 84.6scoreReported result | Not published |
| CharXiv Reasoningwith code interpreter | Not published | 90.6scoreReported result | Not published |
| Price per million tokens | |||
| Input | $0.216 | $0.20 (best) | $0.20 (best) |
| Output | $0.647 | $0.50 (best) | $0.75 |
| Cached input | Not published | Not published | $0.02 |
| Blended (3:1) | $0.323 | $0.275 (best) | $0.338 |
| Long-context rate | Same rate | Same rate | Same rate |
| Price source | Median of 15 providers | Median of 5 providers | Official Inception API |
| Limits | |||
| Context window | 1,000,000 tokens (best) | 262,144 tokens | 260,000 tokens |
| Max input | Not published | Not published | Not published |
| Max output | 384,000 tokens (best) | 131,072 tokens | 65,536 tokens |
| Inputs and features | |||
| Text | Yes | Yes | Yes |
| Images | Yes | Yes | No |
| PDFs | No | No | No |
| Audio | No | No | No |
| Video | No | Yes | No |
| Reasoning | Yes | Yes | Yeslow · medium · high |
| Tool calling | Yes | Yes | Yes |
| Structured output | Not published | Not published | Yes |
| File attachments | Yes | Yes | No |
| Temperature control | Yes | Yes | Yes |
| Reasoning control | always | always | effort |
| Availability | |||
| Weights | OpenMIT | Openqwen-community-1.0 | Proprietary |
| API model ID | Not published | Not published | mercury-2.5 |
| API providers | 16 (best) | 6 | 1 |
| Released | Aug 21, 2026 | Aug 27, 2026 | Sep 8, 2026 |
| Knowledge cutoff | Not published | Not published | Not published |
| Publisher documentation | Model weights | Model weights | Publisher documentation |
| Source metadata updated | Sep 1, 2026 | Aug 27, 2026 | Sep 8, 2026 |
| Data sources | models.dev · Epoch AI | models.dev · Epoch AI | models.dev · Epoch AI |
| Catalog checked | Nov 22, 2026 | Nov 22, 2026 | Nov 22, 2026 |
What would a month cost?
Enter your expected volume in millions of tokens. List prices only; caching and batch discounts would lower these.
DeepSeek V4 Flash Vision Exp$3.45
Qwen3.8 Flash Next$3.00
Mercury 2.5$3.50
Which should you choose?
Which is better: DeepSeek V4 Flash Vision Exp, Qwen3.8 Flash Next or Mercury 2.5?
There is not enough shared capability evidence to name an overall winner. The scores (DeepSeek V4 Flash Vision Exp 68/100, Qwen3.8 Flash Next 65/100, Mercury 2.5 62/100) compare the available specifications and price. The score weighs price 71%, context window 29%. None of the shared benchmarks cover every model here yet, so capability is left out of this verdict.There is not enough shared capability evidence to name an overall winner. The scores (DeepSeek V4 Flash Vision Exp 64/100, Qwen3.8 Flash Next 43/100, Mercury 2.5 43/100) compare the available specifications only. The score weighs context window 100%. None of the shared benchmarks cover every model here yet, so capability is left out of this verdict.
Which is cheaper, DeepSeek V4 Flash Vision Exp, Qwen3.8 Flash Next or Mercury 2.5?
Qwen3.8 Flash Next is cheaper at $0.20 input / $0.50 output per million tokens (median across 5 API providers). DeepSeek V4 Flash Vision Exp costs $0.216 input / $0.647 output per million tokens (median across 15 API providers); Mercury 2.5 costs $0.20 input / $0.75 output per million tokens (official Inception API price). At a typical mix of three input tokens to one output token, that is $0.275 per million tokens for Qwen3.8 Flash Next versus $0.323 for DeepSeek V4 Flash Vision Exp (1.2× as much) and $0.338 for Mercury 2.5 (1.2× as much).
Which scores higher on benchmarks?
There is no independent benchmark that covers all three models yet. DeepSeek V4 Flash Vision Exp has not been scored yet, Qwen3.8 Flash Next has not been scored yet and Mercury 2.5 has not been scored yet.
Which is better for coding?
There are no published SWE-bench Verified results for DeepSeek V4 Flash Vision Exp, Qwen3.8 Flash Next and Mercury 2.5 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?
DeepSeek V4 Flash Vision Exp has the largest context window at 1,000,000 tokens, against 262,144 for Qwen3.8 Flash Next and 260,000 for Mercury 2.5. Maximum output per response: DeepSeek V4 Flash Vision Exp up to 384,000, Qwen3.8 Flash Next up to 131,072, Mercury 2.5 up to 65,536 tokens.
Which can read images, PDFs, audio or video?
DeepSeek V4 Flash Vision Exp accepts text and images; Qwen3.8 Flash Next accepts text, images and video; Mercury 2.5 accepts text. Qwen3.8 Flash Next handles the widest range of inputs.
Are any of these open source?
DeepSeek V4 Flash Vision Exp and Qwen3.8 Flash Next publishes its weights (MIT and qwen-community-1.0) and can be self-hosted; Mercury 2.5 is proprietary.
Which is newer?
Mercury 2.5 is the newest, released Sep 8, 2026. Qwen3.8 Flash Next came out Aug 27, 2026; DeepSeek V4 Flash Vision Exp came out Aug 21, 2026.
How do you decide the winner?
This rating uses price (71%, blended price per million tokens (3 input : 1 output), log scale); context window (29%, 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 Nov 22, 2026.This rating uses context window (100%, 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 Nov 22, 2026.