Mercury 2.5 vs Claude Haiku 5.5 vs DeepSeek V4 Flash Vision Exp
Claude Haiku 5.5 comes out ahead, 76 to 65 and 55 on our weighted score, and it is the cheaper option too.
InceptionMercury 2.5
55/100- ECI—
- Price$0.20 / $0.75
- Context260K
- Our pick
AnthropicClaude Haiku 5.5
76/100- ECI160.4
- Price$0.10 / $0.50
- Context1M
DeepSeekDeepSeek V4 Flash Vision Exp
65/100- ECI—
- Price$0.216 / $0.647
- Context1M
Claude Haiku 5.5 is our pick
Claude Haiku 5.5 is the better all-round choice, scoring 76/100 against DeepSeek V4 Flash Vision Exp (65) and Mercury 2.5 (55). It leads on price and inputs & features. 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 priceClaude Haiku 5.5Claude Haiku 5.5 $0.20 · DeepSeek V4 Flash Vision Exp $0.323 · Mercury 2.5 $0.338 per 1M tokens (3:1 blend)
- Longest contextClaude Haiku 5.5 and DeepSeek V4 Flash Vision ExpClaude Haiku 5.5 1,000,000 · DeepSeek V4 Flash Vision Exp 1,000,000 · Mercury 2.5 260,000 tokens
- Widest inputsClaude Haiku 5.5Mercury 2.5: Text · Claude Haiku 5.5: Text, Images, PDFs · DeepSeek V4 Flash Vision Exp: Text, Images
- Self-hostingDeepSeek V4 Flash Vision ExpPublishes downloadable weights (MIT)
| Measure | Weight | Mercury 2.5 | Claude Haiku 5.5 | DeepSeek V4 Flash Vision Exp |
|---|---|---|---|---|
| Price | 50% | 69 | 80 | 70 |
| Inputs & features | 30% | 45 | 80 | 60 |
| Context window | 20% | 36 | 60 | 60 |
| Overall | 100% | 55/100 | 76/100 | 65/100 |
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 | Mercury 2.5Inception | Claude Haiku 5.5Anthropic | DeepSeek V4 Flash Vision ExpDeepSeek |
|---|---|---|---|
| Capability | |||
| Capabilities Index (ECI) | — | 160.4 | — |
| ECI rank | — | #14 of 165 | — |
| GPQA DiamondGraduate-level science questions | — | 89.7% | — |
| FrontierMath Tiers 1–3Research-level mathematics | — | 75.1% | — |
| OTIS Mock AIME 2024–2025Competition mathematics | — | 98.9% | — |
| SimpleQA VerifiedShort factual questions | — | 23.8% | — |
| Price per million tokens | |||
| Input | $0.20 | $0.10 (best) | $0.216 |
| Output | $0.75 | $0.50 (best) | $0.647 |
| Cached input | $0.02 | $0.01 (best) | — |
| Blended (3:1) | $0.338 | $0.20 (best) | $0.323 |
| Long-context rate | Same rate | Over 100K: $0.50 / $2.50 | Same rate |
| Price source | Official Inception API | Official Anthropic API | Median of 14 providers |
| Limits | |||
| Context window | 260,000 tokens | 1,000,000 tokens (best) | 1,000,000 tokens (best) |
| Max output | 65,536 tokens | 128,000 tokens | 384,000 tokens (best) |
| Inputs and features | |||
| Text | Yes | Yes | Yes |
| Images | No | Yes | Yes |
| PDFs | No | Yes | No |
| Audio | No | No | No |
| Video | No | No | No |
| Reasoning | Yeslow · medium · high | Yeslow · medium · high · xhigh · max | Yes |
| Tool calling | Yes | Yes | Yes |
| Structured output | Yes | Yes | No |
| Availability | |||
| Weights | Proprietary | Proprietary | OpenMIT |
| API model ID | mercury-2.5 | claude-haiku-5-5 | — |
| API providers | 6 | 33 (best) | 15 |
| Released | Sep 8, 2026 | Oct 7, 2026 | Aug 21, 2026 |
| Knowledge cutoff | — | Jun 2026 | — |
What would a month cost?
Enter your expected volume in millions of tokens. List prices only; caching and batch discounts would lower these.
- Mercury 2.5$3.50
- Claude Haiku 5.5$2.00
- DeepSeek V4 Flash Vision Exp$3.45
Which should you choose?
Which is better: Mercury 2.5, Claude Haiku 5.5 or DeepSeek V4 Flash Vision Exp?
Claude Haiku 5.5 is the better all-round choice, scoring 76/100 against DeepSeek V4 Flash Vision Exp (65) and Mercury 2.5 (55). It leads on price and inputs & features. 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, Mercury 2.5, Claude Haiku 5.5 or DeepSeek V4 Flash Vision Exp?
Claude Haiku 5.5 is cheaper at $0.10 input / $0.50 output per million tokens (official Anthropic API price). DeepSeek V4 Flash Vision Exp costs $0.216 input / $0.647 output per million tokens (median across 14 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 $0.20 per million tokens for Claude Haiku 5.5 versus $0.323 for DeepSeek V4 Flash Vision Exp (1.6× as much) and $0.338 for Mercury 2.5 (1.7× as much).
Which scores higher on benchmarks?
There is no independent benchmark that covers all three models yet. Mercury 2.5 has not been scored yet, Claude Haiku 5.5 has an ECI of 160.4 and DeepSeek V4 Flash Vision Exp has not been scored yet.
Which is better for coding?
There are no published SWE-bench Verified results for Mercury 2.5, Claude Haiku 5.5 and DeepSeek V4 Flash Vision Exp 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?
Claude Haiku 5.5 and DeepSeek V4 Flash Vision Exp have the largest context windows (1,000,000 and 1,000,000 tokens), against 260,000 for Mercury 2.5. Maximum output per response: Mercury 2.5 up to 65,536, Claude Haiku 5.5 up to 128,000, DeepSeek V4 Flash Vision Exp up to 384,000 tokens.
Which can read images, PDFs, audio or video?
Mercury 2.5 accepts text; Claude Haiku 5.5 accepts text, images and PDFs; DeepSeek V4 Flash Vision Exp accepts text and images. Claude Haiku 5.5 handles the widest range of inputs.
Are any of these open source?
DeepSeek V4 Flash Vision Exp publishes its weights (MIT) and can be self-hosted; Mercury 2.5 and Claude Haiku 5.5 is proprietary.
Which is newer?
Claude Haiku 5.5 is the newest, released Oct 7, 2026. Mercury 2.5 came out Sep 8, 2026; DeepSeek V4 Flash Vision Exp came out Aug 21, 2026. Knowledge cutoff: Claude Haiku 5.5 Jun 2026.
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 Nov 2, 2026.
